Paddock moving image division system

Abstract

To provide a paddock moving image division system that receives a moving image of a paddock at a horse race place, and can divide a single moving image in each horse by determining a horse number displayed into the moving image.SOLUTION: A paddock moving image division system according to the present invention, includes an image processing device that receives a moving image of a paddock of a horse race place and divides them into the moving image of each race horse. The image processing device comprises an AI that learns a house number displayed on a telop displayed in the moving image as a first feature point, and determines the horse number by the AI from the moving image to be received, and determines a position that the moving image is divided by determining a time when the other race horse is not reflected while enabling a single display of each race horse in each race horse on the basis of the horse number.SELECTED DRAWING: Figure 1

Description

[Technical field]

[0001] The present invention relates to a paddock video division system, and in particular to a paddock video division system that receives video of a racecourse paddock, identifies the horse numbers displayed in the video, and enables division into individual videos for each horse. [Background technology]

[0002] At racetracks, the condition of each racehorse can be inspected in the paddock before the race starts. The condition of the racehorses walking in the paddock can be viewed as a video of the paddock on a website provided by the organizer of the horse race, without the need to go to the paddock to see it. This video of the paddock is made up of a series of consecutive images of each racehorse walking in the paddock, so viewers must manually find the position of the horse they want to see.

[0003] Racehorses usually appear in the paddock 30 minutes before the start of the race, and if you are buying a winning ticket, you need to determine the condition of the horse before the race during this time. If you want to compare the current condition of the horse with its past condition to determine its current condition, it is possible to view past videos. However, in order to do this, you need to find the position in the paddock video where each horse is displayed, and also launch each past video to compare them, which requires quick operation in the short time before the race. In order to make it easier to check the condition of the horses, methods have been proposed to make paddock video easier to watch.

[0004] Patent document 1 discloses a horse information distribution system that includes multiple image acquisition units that acquire images from the normal direction of the paddock circular track, a panoramic image creation unit that arranges the images acquired by each image acquisition unit to create a panoramic image, and a single-horse image creation unit that trims the panoramic image so that the horse's body is positioned in a fixed position. According to the invention described in Patent Document 1, videos of racehorses walking in the paddock are trimmed to create standalone images of each racehorse, allowing users of this horse information distribution system to select the racehorse they desire and obtain that racehorse's video and related information.

[0005] On the other hand, the system described in Patent Document 1 uses a method of analyzing images of bibs showing horse numbers to identify racehorses from video of the paddock, and therefore requires multiple cameras along the circuit to reliably capture images of bibs showing horse numbers. However, since there are multiple racecourses in various locations, it is not easy to install multiple cameras along the circuit in the paddocks of all of these racecourses.

[0006] For example, in horse racing organized by the Japan Racing Association (JRA), the paddock video provided by the Japan Racing Association is the only official video, and in order to provide a method for making the paddock video easier to view, it is necessary to base it on such existing official videos.Therefore, it is desirable to provide a system that, while based on the existing official videos, makes the paddock video easier to view in the limited time between the distribution of the paddock video and before the race, and makes it easier to understand by adding related information. [Prior art documents] [Patent documents]

[0007] [Patent Document 1] JP 2021-166383 A Summary of the Invention [Problem to be solved by the invention]

[0008] The present invention has been made in consideration of the problems with the conventional paddock video provision technology described above, and an object of the present invention is to provide a paddock video splitting system that receives video of a racecourse paddock, identifies the horse numbers displayed within the video, and makes it possible to split the video into individual videos for each horse. [Means for solving the problem]

[0009] The paddock video splitting system of the present invention, made to achieve the above-mentioned object, is a video splitting system having an image processing device which receives video of the paddock at a racecourse and splits it into videos for each racehorse, and the image processing device is equipped with an AI which learns the horse number displayed in the subtitles displayed within the video as a first feature point, and determines the horse number from the received video using the AI, and based on the horse number, determines the position at which to split the video by determining the time when each racehorse can be displayed alone and other racehorses are not visible.

[0010] It is preferable that the AI ​​of the image processing device further learns the horse numbers displayed on the bibs worn by each racehorse in the video as a second feature point, and the image processing device determines the position to divide the video based on the first feature point and the second feature point. It is preferable that the image processing device discriminates the first feature point and the second feature point from the video, and based on the result of the discrimination, determines the positions on the video where the display of each racehorse begins and ends from the time when each racehorse is displayed alone, and at least one of dividing the video based on the positions on the video, or adding the positions on the video where the display of each racehorse begins and ends as playback points to the video without dividing the video.

[0011] When dividing the moving image, the image processing device preferably creates each divided moving image as a moving image file for each racehorse separately from the original moving image. It is preferable that the image processing device preferentially uses the first feature point to determine the positions on the video where the display of each racehorse begins and ends, and uses the second feature point to determine the positions when the first feature point and the second feature point are distinguished and the respective horse numbers differ, or when the first feature point cannot be distinguished.

[0012] Videos obtained by dividing the video based on positions on the video, or videos to which the positions on the video where the display starts and ends for each racehorse are added as playback points, are provided for viewing on a website provided by the image processing device, and it is preferable that the videos provided for viewing on the website are provided with the option of at least one of time-specified fast-forwarding or rewinding, double speed or slow speed, low image quality or high image quality, and loop playback. It is preferable that the videos provided for viewing on the website be provided in such a way that a user can select a past paddock video of the racehorse displayed in the video or of another racehorse and compare the video with the past paddock video.

[0013] The videos that are made available for viewing on the website are preferably provided free of charge after logging in with a member-only ID registered on the website, or in response to events, campaigns, and the like. It is preferable that the image processing device accepts input of a manual splitting instruction by an operator, and upon accepting the input of the manual splitting instruction, further accepts input of a designation of a display start or display end at an arbitrary video playback position by the operator, and splits the video based on the specified display start or display end position.

[0014] It is preferable that the image processing device is equipped with an acquisition program that acquires information about horse race events, including at least one of the following provided by the race organizer: the name of the race venue, the date of the race, the horse's name, the race results, and withdrawals from the race, and that the acquisition program manually or automatically imports the information about the horse race events and provides it for viewing on the website together with video footage of the paddocks for each racehorse. Effect of the Invention

[0015] According to the paddock video splitting system of the present invention, after receiving official paddock video streamed in real time, the horse numbers are immediately identified using AI and individual images are created for each racehorse, making it possible to provide individual images for each racehorse created in a relatively short period of time.

[0016] Furthermore, according to the paddock video splitting system of the present invention, the horse numbers displayed in the captions displayed within the paddock video are identified and individual images for each racehorse are generated, so that even in situations where the horse numbers shown on the bibs of each racehorse are not photographed in a way that makes them easy to identify, the video can be appropriately divided in a short amount of time based on the horse numbers displayed in the captions.

[0017] Furthermore, according to the paddock video splitting system of the present invention, if the horse number shown on the bib of each racehorse is captured in such a way that it can be distinguished, the horse number shown on the bib is distinguished in addition to the horse number displayed in the caption, and if the two horse numbers match, the horse number that matches is deemed correct and the horse number displayed in the caption is adopted as is, and if it is determined that the horse number displayed in the caption and the horse number shown on the bib are different, the horse number shown on the bib is given priority, thereby ensuring the accuracy of identifying racehorses. Also, even if the caption is not displayed due to some malfunction, it is possible to identify each racehorse from the horse number shown on the bib, improving the accuracy and probability of identifying racehorses. [Brief description of the drawings]

[0018] [Figure 1] 1 is a block diagram showing an outline of the overall configuration of a paddock video splitting system according to an embodiment of the present invention. [Diagram 2] 1 is a diagram illustrating an example of one screen of a paddock video input to a paddock video division system according to an embodiment of the present invention. [Diagram 3] 1 is a diagram for explaining an AI learning and judgment method of an image processing device according to an embodiment of the present invention. [Figure 4] 1 is a diagram illustrating a horse number determination method according to an embodiment of the present invention. [Diagram 5] 1 is a diagram illustrating a horse number determination method according to an embodiment of the present invention. [Figure 6] 1 is a diagram illustrating an overall processing flow of a paddock video splitting system according to an embodiment of the present invention. [Figure 7] FIG. 2 is a diagram for explaining in outline parallel processing by AI in an image processing device according to an embodiment of the present invention. [Figure 8] 1 is a flowchart illustrating a paddock video division method according to an embodiment of the present invention. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0019] Next, a specific example of an embodiment for implementing the paddock video division system according to the present invention will be described in detail with reference to the drawings. FIG. 1 is a block diagram showing an outline of the overall configuration of a paddock video division system according to an embodiment of the present invention.

[0020] 1, a paddock video splitting system 1 according to an embodiment of the present invention has an image processing device 10 and a user terminal 20, which are connectable via a wide area network 5 such as the Internet. The image processing device 10 is a device that receives video of the paddock at a racecourse and splits it into videos for each racehorse and provides them, so that a horse racing fan user can access a website provided by the image processing device 10 with the user terminal 20 and select and view the video he or she wants to view from the videos split for each racehorse.

[0021] At the racetrack, racehorses are led to the paddock before the start of the race and walk around the paddock. By observing the racehorses walking around the paddock, horse racing fans can judge the physical condition of the racehorses just before the race. The appearance of the racehorses walking around the paddock is provided as one continuous video by the organizer of the horse race not only to horse racing fans present at the racetrack but also to horse racing fans watching the horse racing broadcast remotely. The video of the paddock provided at this time provides images of each racehorse evenly so that there is no bias towards the racehorses walking around the paddock, so it is not necessarily convenient for horse racing fans who want to know information about a specific racehorse.

[0022] The paddock video splitting system 1 according to an embodiment of the present invention receives paddock video provided by the organizer of the horse race, identifies the horse numbers displayed in the received paddock video, and splits and provides videos for each racehorse based on the identified horse numbers, allowing users who are horse racing fans to select the video of the racehorse they want to see and watch it with concentration.

[0023] As shown in FIG. 1, the image processing device 10 has a control unit 11, an input / output unit 12, a memory unit 13, a display unit 14, a member registration management unit 15, an image discrimination unit 16, an image division unit 17, a related information acquisition unit 18, and an image provision unit 19.

[0024] The member registration management unit 15 registers and manages users who use the paddock video splitting system 1 as user members. A user who is going to use the paddock video splitting system 1 for the first time accesses the website provided by the image processing device 10 with the user terminal 20, and enters user information required for registration, such as name, email address, and a password required for accessing member-only pages from the next time onwards, on the new member registration web page, to register as a member.

[0025] When the member registration management unit 15 accepts the input of user information required for registration, it issues a user ID unique to the user, associates the accepted user information with the issued user ID, and stores and manages them as user membership information in the storage unit 13. In addition, the member registration management unit 15 accepts and authenticates the user ID and password entered by a registered user member from the next time onwards, and permits access to the members-only page if the combination of the user ID and password matches the registered user membership information.

[0026] The image discrimination unit 16 is equipped with an AI that learns the horse numbers included in the captions displayed in the video of the paddock provided by the organizer of the horse race as a first feature point, and this AI discriminates the horse numbers of the racehorses displayed in the video of the paddock. The AI ​​equipped in the image discrimination unit 16 further learns the horse numbers displayed on the bibs worn by each racehorse in the video of the paddock as a second feature point, and when the horse number on the bib of the racehorse displayed centrally is captured in the provided video of the paddock in a state where it can be confirmed, the AI ​​further discriminates the horse number on the bib.

[0027] In the paddock video provided in this way, there may be a case where a first horse number determined from a horse number included in a caption and a second horse number determined from a horse number displayed on a race bib are present. In this case, the image discrimination unit 16 judges the first horse number as the correct horse number when only the first horse number is discriminated, judges the matching horse number as the correct horse number when the discriminated first horse number and the second horse number match, judges the second horse number as the correct horse number when only the second horse number is discriminated, and judges the second horse number as the correct horse number when the discriminated first horse number and the second horse number do not match.

[0028] The reason why the second horse number is preferentially determined to be the correct horse number is that the horse number included in the caption may contain human errors, whereas the horse number displayed on the bib is almost never mistaken. The reason why the AI ​​equipped in the image discrimination unit 16 learns the horse number included in the caption as the first feature point is that, whereas captions are displayed stably in fixed positions in the video of the paddock provided by the organizer of the horse race, bib numbers may be affected by people or other racehorses and are not necessarily displayed stably and identifiable. Here, it has been described that one AI learns the first feature points and the second feature points, but in other embodiments, multiple AIs may be provided, and the AI ​​that learns the first feature points and the AI ​​that learns the second feature points may be configured separately.

[0029] The image discrimination unit 16 may discriminate the horse number for each scene while playing back the provided paddock video and continue to output the horse number that it has determined to be the correct horse number, or it may associate the section of the video within which the correct horse number was determined by discriminating within the paddock video with the correct horse number determined at that time, and store these in the memory unit 13.

[0030] An image dividing unit 17, based on the horse number determined by the image determining unit 16, determines the position to divide the provided paddock video by determining the time when each racehorse can be displayed alone and other racehorses are not shown. In one embodiment, the image division unit 17 adds the start and end points of the time during which each racehorse can be displayed alone and other racehorses are not shown to the paddock video provided as video playback points for that racehorse to the base paddock video. This makes it possible to display and play back each racehorse by specifying the playback points using the video playback function or on the system side.

[0031] In another embodiment, the image division unit 17 divides the provided paddock video into individual videos at the start and end points of the time when each racehorse can be displayed alone and other racehorses are not shown. The divided videos are stored in the memory unit 13 as videos separate from the original paddock video. If the video portion of the horse number identified by the image discrimination unit 16 exists in multiple places in the original paddock video, thereby generating multiple divided videos for the same horse number, the image division unit 17 associates these multiple divided videos and stores them in the memory unit 13. In this case, the file names of the divided videos may be given the race date, a code that uniquely identifies the racecourse, the race number, and the corresponding horse number to identify the race. This makes it possible to determine the content of the divided videos from the file names.

[0032] The image division unit 17 also accepts input of a manual division instruction from the manager, and upon accepting the input of the manual division instruction, further accepts input of a designation of a display start or display end at an arbitrary video playback position by the manager, and divides the video based on the designated display start or display end position. This makes it possible to generate a divided video even if the video provided is a video in a situation where neither subtitles nor bib numbers can be confirmed, and therefore the video cannot be automatically divided by the image processing device 10. The image division unit 17 may mark and save the video section in the original paddock video that cannot be automatically divided because the horse number cannot be confirmed, even in a scene that is effective for creating a divided video, so that the manager can know.

[0033] The related information acquisition unit 18 acquires related information to be added to the video when the video with the added playback points or the divided video, determined by the image division unit 17, is provided to user members via a website after determining the division positions. The related information includes the name of the racecourse, the date of the race, the horse's name, the race results, and withdrawals, which are provided by the race promoter, and the acquired related information does not need to include all of this information, but includes at least one of the name of the racecourse, the date of the race, the horse's name, the race results, and withdrawals.

[0034] The related information acquisition unit 18 includes an acquisition program for acquiring information on the horse race event, and starts the acquisition program to acquire related information related to the provided paddock video from a website provided by the horse race event organizer. In addition, if the image processing device 10 stores past divided videos and the like as a database, the acquisition program may be configured to search the database and extract the necessary related information.

[0035] The image providing unit 19 determines the positions at which the image is divided by the image dividing unit 17, and provides the video with the added playback points or the divided video in combination with related information related to this video acquired by the related information acquiring unit 18 to the user member via the website. As described above, when there are multiple divided videos for the same horse number, the image providing unit 19 connects these multiple divided videos or plays them consecutively, and provides them so that the user member can view them all at once.

[0036] The videos provided by the image providing unit 19 are provided in the form of thumbnails of all racehorses participating in the same race on the video selection web page, combined with a title that combines the horse number and the horse name, so that user members can select the video of the racehorse they wish to view and easily view the selected video on an enlarged screen.

[0037] The image providing unit 19 provides a selection button on the web page to select either a single video or a comparative video so that when a user member selects and views a split image of a specific racehorse, the user member can view the image while comparing it side by side with a past paddock video of the selected racehorse, or side by side with a current or past paddock video of another racehorse different from the selected racehorse. When the user member selects a comparative video, the image providing unit 19 further provides a selection menu for selecting which racehorse and when to compare the data, thereby enabling the user member to view the image while comparing it with the desired comparative video. In addition, the image providing unit 19 provides at least one of the following selectable options for time-specified fast-forward or rewind, double or slow speed, low or high image quality, and loop playback, for both a single video and a comparison video, thereby improving the usability for user members when viewing videos.

[0038] The video provided by the image providing unit 19 is provided within a member-only page that can be viewed by user members by logging in, but the image providing unit 19 accepts input of an instruction for free provision of the video for events, campaigns, etc. by an administrator, and upon receiving the input of the instruction for free provision, provides the video so that it can be viewed within a Web page that does not require login.

[0039] The input / output unit 12 includes communication means for receiving paddock videos provided by the horse race organizer, receiving video viewing requests from the user terminal 20 via a website and transmitting the corresponding divided videos. The input / output unit 12 also includes input means such as a keyboard and mouse for the operator to input instructions for manual division and instructions for providing videos free of charge.

[0040] The storage unit 13 stores video data such as the provided paddock video and divided videos, data such as user information of the user group committee members, and a control program for controlling the image processing device 10 itself. A database of past video data and the like is also stored in the storage unit 13. The storage unit 13 is realized by a hard disk device, a semiconductor memory device, or the like.

[0041] The display unit 14 displays the video data and user information stored in the storage unit 13, as well as setting screens for events, campaigns, etc. In addition, it displays programs such as the control program for the image processing device 10 as necessary. The display unit 14 is realized by a display device such as a liquid crystal display. The control unit 11 controls each component of the image processing device 10 to perform the functions described above in accordance with a control program stored in the storage unit 13. The control unit 11 is realized by a control circuit including a semiconductor device such as a microcomputer.

[0042] The user terminal 20 includes a control unit 21, an input / output unit 22, a storage unit 23, and a display unit 24. The input / output unit 22 includes a communication means for transmitting user information, a request to view a video, and the like to the image processing device 10, and for receiving divided videos from the image processing device 10. The input unit 22 also includes an input means such as a keyboard, a mouse, or a touch panel for inputting user information and selecting divided videos provided on a website provided by the image processing device 10. The storage unit 23 temporarily stores video data provided by the image processing device 10 and stores programs such as a control program for the user terminal 20.

[0043] The display unit 24 displays a selection screen for video data provided by the image processing device 10, and displays video data received as a result of the selection. The control unit 21 controls each component of the user terminal 20 to perform the functions described above in accordance with the control program stored in the storage unit 23. The user terminal 20 is realized by a personal computer, a tablet terminal, a smartphone, or the like.

[0044] FIG. 2 is a diagram illustrating an example of one screen of a paddock video input to a paddock video division system according to an embodiment of the present invention. The paddock video provided to the paddock video division system 1 is received by the image processing device 10 as an input video 30, and referring to Fig. 2, the input video 30 includes a racehorse 31 walking around the paddock with a bib number 33 attached. The number "2" assigned to the racehorse 31 for the race is attached to the bib number 33 as a horse number 35 of the bib 33.

[0045] Additionally, captions 32 are displayed in two places, above and below, corresponding to racehorse 31, with the upper caption 32 displaying the race number and the lower caption 32 displaying information related to racehorse 31. The number "2" is also displayed in the lower caption 32 as horse number 34 of the caption 32. The lower caption 32 also includes the name of racehorse 31, "Umaniban," its weight, "460 kg," and the name of its jockey, "Jockey Ichiro."

[0046] 2, the input video 30 includes a racehorse 31 by itself, but the input video 30 does not necessarily include a racehorse 31 by itself, and there may be cases where previous and next racehorses 31 are also shown. Therefore, when the image division unit 17 divides the video, it determines and divides the time when each racehorse 31 can be displayed by itself and other racehorses 31 are not shown.

[0047] FIG. 3 is a diagram illustrating an AI learning and judgment method of an image processing device according to an embodiment of the present invention. Referring to FIG. 3, the AI ​​40 included in the image discrimination unit 16 employs two different AI algorithms: deep learning and template matching.

[0048] In deep learning, the probability of identification is improved by having the AI ​​40 learn images of the target image in various states. In the example of Fig. 3, the number "3" is input as input image 41. The number "3" in various shapes and colors is input in advance to the AI ​​40 as learning pattern images 42, and the AI ​​40 is made to learn what kind of input is the number "3". This allows the AI ​​40 to identify that the character input as input image 41 is "3".

[0049] On the other hand, when the characters to be identified are limited in advance, such as identifying a horse number, template matching can also be used. In the example of Fig. 3, the numbers 1 to 6 are individually prepared as template images 44. When the number "4" is input as the input image 43, the AI ​​40 performs a pattern comparison with the template image 44 and is able to identify the number "4" with the highest degree of pattern match.

[0050] In an embodiment of the present invention, the AI ​​40 identifies the horse number for each image of the paddock video provided as input by both deep learning and template matching, and determines that the horse number is correct as long as the horse number identified by deep learning and the horse number identified by template matching continue to match. If the horse number identified by deep learning and the horse number identified by template matching do not match, the AI ​​40 determines that the image is one in which the horse number cannot be identified.

[0051] In another embodiment, when a mismatch occurs, the two numbers identified by the respective AI algorithms may be temporarily stored as horse number candidates, and one of the temporarily stored horse number candidates may be determined to be the correct horse number based on other environmental conditions such as the presence or absence of discontinuous switching of the video. The video of the paddock that is usually provided is continuously shot for a period of time that allows the walking state of the racehorse 31 to be confirmed, so the video does not switch from one to the next in a very short time. Therefore, based on the premise that the horse number does not change while there is no discontinuous switching of the video such as switching to the image of the next racehorse 31, if the temporarily stored horse number candidates include a horse number that was stably identified as the correct horse number immediately before, this horse number may be determined to be the correct horse number.

[0052] Although FIG. 3 shows one AI 40 performing two different AI algorithms, deep learning and template matching, in other embodiments, the AI ​​performing classification using deep learning and the AI ​​performing classification using template matching may be configured as separate AIs 40.

[0053] FIG. 4 is a diagram illustrating a horse number determination method according to an embodiment of the present invention. In FIG. 4, the input video 30 on the left shows a racehorse 31 with the horse number 35 on the bib 33 being "2", and the caption 32 shows that the horse number 34 on the caption 32 is "2". In this case, the AI ​​40 first checks the horse number 34 on the caption 32, which is the first feature point, and determines that the horse number 34 on the caption 32 is "2". Next, since the horse number 35 on the bib 33 is clearly visible in the input video 30, the AI ​​40 checks the horse number 35 on the bib 33, which is the second feature point, and determines that the horse number 35 on the bib 33 is "2". This distinguishes the first horse number from the second horse number, and since the two horse numbers match as a result of the distinction, the image distinction unit 16 determines that "2" is the correct horse number. Therefore, after the input video 30 in FIG. 4 is divided, it is saved as a divided image of the horse number "2".

[0054] FIG. 5 is a diagram illustrating a horse number determination method according to an embodiment of the present invention. In Fig. 5, the input video 30 on the left shows a racehorse 31 with horse number 35 on bib 33 of "3", and the caption 32 shows that horse number 34 on caption 32 is "2". In this case, AI 40 first determines that horse number 34 on caption 32, which is a first feature point, is "2", and then determines that horse number 35 on bib 33, which is a second feature point, is "3". As a result of the determination, the two horse numbers do not match, so image discrimination unit 16 gives priority to horse number 35 on bib 33, which is the second feature point, and determines that the horse number of this input video 30 is "3". Therefore, after being divided, input video 30 in Fig. 5 is saved as a divided image of horse number "3".

[0055] It should be noted that the information in the caption 32 shown in the input video 30 remains the original information regarding the horse number "2" even after splitting, so in one embodiment, in the video provided as a split video, the information regarding the horse number "2" in the caption 32 is blurred to prevent erroneous information from being displayed, or the information regarding the horse number "2" is overwritten with correct information regarding the horse number "3" so as to cover it up, thereby matching the horse number 35 on the bib number 33 with the information in the caption 32.

[0056] FIG. 6 is a diagram illustrating an outline of the overall flow of processing in the paddock video division system according to the embodiment of the present invention. Before the start of a race, a paddock video division system 1 according to an embodiment of the present invention receives paddock video provided by the organizer of the horse race as an input video 30. The AI ​​40 included in the image processing device 10 determines the horse number for each image or scene of the received input video 30 based on the first feature point and / or the second feature point, and determines a division point 50, which is the position at which to divide the input video 30, by determining a time during which each racehorse 31 can be displayed alone and no other racehorses 31 are shown within a section of the input video 30 in which the correct horse number has been determined by the image division unit 17.

[0057] Next, the image division unit 17 divides the input video 30 according to the division points 50 to generate divided videos 51. The divided videos 51 are generated by extracting video portions that satisfy the conditions that the correct horse numbers are identified within the original input video 30, and that each racehorse 31 can be displayed independently and that other racehorses 31 are not shown.

[0058] When the divided video 51 is generated, the related information acquisition unit 18 launches a related information acquisition program and acquires related information 52 related to the divided video 51 from a website provided by the horse race organizer or a database provided in the image processing device 10 that stores past divided videos 51, etc.

[0059] When there are multiple divided videos 51 with the same horse number in the input video 30, the image providing unit 19 groups these multiple videos by horse number, generates an output video 54 by combining it with related information 52 acquired for each horse number, and provides it to the user member so that it can be viewed via a website 53 provided by the image processing device 10. The user member can select and view the provided output video 54 by logging into the website 53 with the user terminal 20 and accessing a member-only page. In the above description, the website 53 is directly provided by the image processing device 10, but in other embodiments, the website 53 may be configured to be provided from a dedicated server separate from the image processing device 10, and the output video 54 output by the image processing device 10 may be provided to the user via the dedicated server.

[0060] FIG. 7 is a diagram for explaining in outline parallel processing by AI in an image processing device according to an embodiment of the present invention. The paddock video that the image processing device 10 receives as the input video 30 is a video of all the racehorses 31 running in the same race walking around, and since there are a maximum of 18 horses in the case of JRA horse racing, for example, generating divided images for each horse in turn would result in a long image division processing time. The paddock video needs to be provided during the limited time from when each racehorse 31 appears in the paddock until the deadline for purchasing winning betting tickets before the race.

[0061] Therefore, the AI ​​40 included in the image processing device 10 according to the embodiment of the present invention is configured to divide the input video 30 in parallel to avoid long image division processing times. The AI ​​140 can read large amounts of data for multiple horses in one process and operate in parallel, regardless of image quality such as HD, 4K, or 8K, to easily process the determination of horse numbers in the captions 32 and bibs 33.

[0062] In the example of Fig. 7, the input video 30 is input to the AI ​​40 as three input videos 30, and the AI ​​40 determines the division points for each of the three horse numbers by parallel processing. When the input video 30 is not divided and is provided from the website 53 as a video with playback points added for the horse numbers, the image processing device 10 generates and outputs a video for each horse number in which the division points for the horse numbers are added to the input video 30 as playback points. When the input video 30 is provided from the website 53 as an output video 54 obtained by dividing the input video 30, the image processing device 10 generates divided videos 51 by dividing the input video 30 based on the division points, and when there are multiple divided videos 51 for the same horse number, the multiple divided videos 51 are collectively generated as an output video 54 together with related information 52, and provided to the user member via the website 53.

[0063] 7 shows three input videos 30 to be processed in parallel, but more or less videos may be input to the AI ​​40. Also, instead of processing multiple input videos 30 in parallel with one AI 40, multiple AIs 40 may be provided and the multiple AIs 40 may be operated simultaneously to perform parallel processing. By performing parallel processing in this manner using one or more AIs 40, the paddock video splitting system 1 according to an embodiment of the present invention is capable of outputting the output video 54 within three minutes of receiving the input video 30.

[0064] FIG. 8 is a flowchart illustrating a paddock video division method according to an embodiment of the present invention. Referring to FIG. 8, in step S800, the image processing device 10 receives video of the paddock provided by the horse race organizer.

[0065] When the image processing device 10 receives the paddock video, it inputs the received paddock video to the AI ​​40 in step S810 and determines the horse number 34 of the caption 32 included in the caption 32 displayed in the paddock video. Furthermore, the image processing device 10 checks whether the horse number 35 on the bib 33 attached to the racehorse 31 is discernible in the image in which the horse number 34 in the caption 32 has been discerned (stage S820), and if the horse number 35 on the bib 33 cannot be confirmed, it determines that the discriminated horse number 34 in the caption 32 is the correct horse number, and in stage S850 determines the time range in the video in which the racehorse 31 is displayed alone based on the horse number 34 in the caption 32.

[0066] On the other hand, if it is possible to confirm the horse number 35 on the bib 33 in step S820, the AI ​​40 determines the horse number 35 on the bib 33 in step S830 and judges whether the horse number 34 on the caption 32 and the horse number 35 on the bib 33 match. If the horse number 34 in the caption 32 matches the horse number 35 on the bib 33, the image processing device 10 determines that the matching horse number is the correct horse number, and in step S850 determines the time range in the video in which the racehorse 31 is displayed alone based on the horse number 34 in the caption 32.

[0067] If the horse number 34 on the caption 32 and the horse number 35 on the bib 33 do not match in step S830, the image processing device 10 determines that the horse number 35 on the bib 33 is the correct horse number, and in step S840 determines the time range in the video in which the racehorse 31 is displayed alone based on the horse number 35 on the bib 33. Finally, in step S860, the image processing device 10 divides the paddock video based on the time range on the video determined in step S840 or step S850.

[0068] If the paddock video is not divided, but the positions on the video where the display of each racehorse 31 begins and ends are added as playback points to the original paddock video, the image processing device 10 adds playback points instead of dividing the video in step S860 and outputs it. The divided video or the paddock video with the added playback points are provided to the user member relatively quickly before the race via the website 53 provided by the image processing device 10. At this time, the user member can select a comparison video as necessary and compare the current video of the racehorse 31 of interest with past videos or with other racehorses 31, thereby making it possible to grasp the current condition of the racehorse 31 in more detail.

[0069] Although the embodiments of the present invention have been described in detail above with reference to the drawings, the present invention is not limited to the above-described embodiments and can be modified in various ways without departing from the technical scope of the present invention. [Explanation of symbols]

[0070] 1. Paddock video splitting system 5. Network 10 Image processing device 11, 21 Control section 12, 22 Input / output section 13, 23 Storage section 14, 24 Display section 15 Membership Registration Management Department 16 Image discrimination section 17 Image division section 18 Related Information Acquisition Department 19 Image provided by 20 User terminal 30 Input video (paddock video) 31 Race Horse 32 Caption 33 Bib Number 34 Horse number in caption 35 Horse number on the race bib 40 AI 41, 43 Input image 42 Learning pattern images 44 Template Images 50 division points 51 Split Video 52 Related Information 53 Websites 54 Output Video

Claims

1. A video splitting system having an image processing device that receives video of a paddock at a racetrack and splits it into videos for each racehorse, The image processing device is equipped with an AI that learns the horse number displayed in the subtitles displayed within the video as a first feature, and the AI ​​determines the horse number from the received video, and based on the horse number, determines the time for each racehorse when each racehorse can be displayed alone and other racehorses are not visible, thereby determining the position to divide the video.

2. The paddock video splitting system described in claim 1, characterized in that the AI ​​of the image processing device further learns the horse numbers displayed on the bibs worn by each racehorse in the video as a second feature point, and the image processing device determines the position to split the video based on the first feature point and the second feature point.

3. The paddock video division system described in claim 2, characterized in that the image processing device discriminates the first feature point and the second feature point from the video, and based on the discrimination result, determines the positions on the video where the display of each racehorse begins and ends from the time when each racehorse is displayed alone, and at least one of dividing the video based on the positions on the video, or adding the positions on the video where the display of each racehorse begins and ends as playback points to the video without dividing the video.

4. The paddock video division system according to claim 3, characterized in that, when dividing the video, the image processing device creates each divided video as a video file for each racehorse separately from the original video.

5. The paddock video division system of claim 3, wherein the image processing device preferentially uses the first feature point to determine the positions on the video where the display of each racehorse begins and ends, and uses the second feature point to determine the positions when the first feature point and the second feature point are distinguished and the respective horse numbers differ, or when the first feature point cannot be distinguished.

6. a video obtained by dividing the video according to a position on the video or a video obtained by adding positions on the video that are the display start and display end of each racehorse as playback points is provided so as to be viewable on a website provided by the image processing device; The paddock video division system according to claim 3 or 4, characterized in that the videos provided for viewing on the website are provided in such a way that at least one of the following can be selected: time-specified fast-forward or rewind, double or slow speed, low or high image quality, and loop playback.

7. The paddock video division system described in claim 6, characterized in that the videos provided for viewing on the website are provided in such a way that past paddock videos of the racehorse displayed in the video or other racehorses can be selected and the video can be compared and played back with the past paddock videos.

8. The paddock video division system according to claim 6, characterized in that the videos provided for viewing on the website are provided free of charge after logging in using a member-only ID registered on the website, or in response to events, campaigns, etc.

9. The paddock video splitting system of claim 3, characterized in that the image processing device accepts input of a manual splitting instruction by an operator, and upon accepting the input of the manual splitting instruction, further accepts input of a designation of a display start or display end at an arbitrary video playback position by the operator, and splits the video based on the specified display start or display end position.

10. The paddock video splitting system described in claim 6, characterized in that the image processing device is equipped with an acquisition program that acquires information about horse race events including at least one of the following provided by the horse racing organizer: the name of the race venue, the date of the race, the horse names, the race results, and / or withdrawals, and the acquisition program manually or automatically imports the horse race event information and provides it for viewing on the website together with paddock video for each racehorse.

Images