JP4452418B2 - How to create a soccer video digest - Google Patents

Description

【００４５】

【００４６】
（７）重要場面の判定
次に、上記映像インデックス（イベント情報）を基に、サッカー番組の重要場面（重要度の高い映像シーン）の判定のために、以下のような重要度判定パラメータを定義する。
【００４７】

【００４８】
上記攻撃レベルの値に、ボール保持レベルパラメータ（Ｐｓ）やボール移動レベルパラメータ（Ｐｂ）を重みをかけて加算することにより、重要度値が算出される。
【００４９】
（８）実施例
前述した構成および動作に基づいて、サッカー映像からダイジェスト映像を作成した。図５および図６の実験データは、２０００年３月１２日のＪリーグ『ヴェルディ川崎対ガンバ大阪』の試合において、 (攻撃レベル＋ボール保持レベルパラメータ）を重要度値としてそれぞれのチームに対して算出した例である。なお、図において、縦軸は演算によって求めた重要度値、横軸は試合の経過の時間軸である。
【００５０】
ところで、上記の重要度値のグラフから、シュートシーンなどの重要場面の判定はある程度可能である。しかし、サッカー番組のダイジェスト生成の場合、重要シーンの判定の他に、そのシーンに至るまでの試合の流れを表現する、ということが重要になってくる。
【００５１】
例えば、重要シーンとして判定されたシーンが複数あった場合、おされ気味の状況においてシュートを放ったシーンであるか、何本もシュートを放つような猛攻撃をかけている際中の１シュートシーンであるかによって、その重要度は自ずと異なってくる。また、そういった試合の流れに関して、「試合の冒頭はホームチームが優勢であったが、次第にアウェイチームが攻め込んできて、先制点を奪ったシーン」であるとか「ずっとホームチームにおされ気味であったが、アウェイチームがようやく攻め込んで１点をとったシーン」といったような内容説明も欲しい。
【００５２】
そこで、本発明では、以下のようにして、試合の流れを考慮した映像シーンの抽出（選択）を行っている。
【００５３】
先ず、この試合の流れを表現するために、以下のような状況レベルを算出する。この状況レベルは、ホーム、アウェイチームそれぞれに算出される。
【００５４】
▲１▼前半、後半をいくつかの区間（時間帯）に分割する。
サッカー試合の場合、通常は、それぞれ約１５分ずつの序盤、中盤、終盤の３つの区間に分割する。
▲２▼上記分割された区間に対して、ボール保持レベルパラメータの積分値（状況レベル）を算出する。
▲３▼積分値がある一定数値以上（＞＝α）であった場合、その区間において、そのチームは攻勢であったと判定する。
▲４▼積分値がある一定数値以下（＞＝β）であった場合、その区間において、そのチームは劣勢であったと判定する。
なお、一例として、上記α、βの値はそれぞれ−１０とする。
▲５▼ホームチーム（Ｈ），アウェイチーム（Ａ）の各チームに対して、それぞれの区間を判定する。
▲６▼結果をすり合わせて、最終的な試合の流れ情報を生成する。
その際の方針は以下の通りである。
Ｈチームが攻勢でＡチームも攻勢 → 激しい攻防
ホームチームが攻勢でアウェイチームも攻勢の場合に、試合の流れ情報として『激しい攻防』を生成する。
Ｈチームが攻勢でＡチームは劣勢 → Ｈ攻勢
ホームチームが攻勢でアウェイチームは劣勢の場合に、試合の流れ情報として『ホームチームが攻勢（Ｈ攻勢）』を生成する。
Ｈチームが劣勢でＡチームは攻勢 → Ａ攻勢
ホームチームが劣勢でアウェイチームは攻勢の場合に、試合の流れ情報として『アウェイチームが攻勢（Ａ攻勢）』を生成する。
Ｈチームが劣勢でＡチームも劣勢 → 両者攻めあぐねる
ホームチームが劣勢でアウェイチームも劣勢の場合に、試合の流れ情報として『両者攻めあぐねる』を生成する。
ただし、「攻めあぐね」と判定された場合でも、その時間帯内にシュート数が一定数以上あったか否か判定し、一定数以上あった場合には、「攻めあぐね」の判定を「攻勢」に変更する。
【００５５】
したがって、試合の流れ情報は、それぞれの時間帯に対して判定された各チームの判定結果の組み合わせによって、ホームチームが攻勢でアウェイチームも攻勢の場合に「激しい攻防」が生成され、ホームチームが攻勢でアウェイチームが劣勢の場合に「ホームチーム攻勢」が生成され、ホームチームが劣勢でアウェイチームが攻勢の場合に「アウェイチーム攻勢」が生成され、ホームチームが劣勢でアウェイチームも劣勢の場合に「両者攻めあぐねる」が生成されることになる。
【００５６】
上記のように各区間が判定されたときに、ダイジェスト作成のシーン抽出において、抽出される重要シーンは上記の判定を反映したものであることが望ましい。そこで以下の抽出ルールによりシーン抽出を行う。
【００５７】
第１に、基本的には、重要度の高い順に選択する。
第２に、重要度が同じシーンが複数あった場合には、激しい攻防、攻勢、攻めあぐねの３つの状態に対してあらかじめ決められているシーン抽出の割合に基づき選択する。
例えば、激しい攻防、攻勢、攻めあぐねの３つの状態に対して、シーン抽出の割合を以下のように決める。
激しい攻防 → Ｈチーム１、Ａチーム１
攻勢 → 攻勢チーム１
攻めあぐね → Ｈチーム0.25、Ａチーム0.25
【００５８】
図５および図６で示した試合に対して、前述した状況レベルを算出し、抽出ルールを適用した結果を以下に示す。
まず状況レベルを算出する。そのためには、ボール保持レベルパラメータの値を、各区間に対して積分する必要がある。図７および図８は各チームのボール保持レベルパラメータの算出結果である。上記の値を３つの区間（序盤、中盤、終盤）に対して、積分した値とその判定結果は表１の通りである。
【００５９】
【表１】

【００６０】
上記により、このサッカー試合の前半は、
序盤：ホームチーム、アウェイチームともに攻めあぐねていたが、
中盤：アウェイチームが攻勢に転じ、
終盤：引き続きアウェイチームが攻勢となる。
という試合の流れであったと表現できる。また、このように試合の流れが明らかになるので、この試合の流れ情報を用いて、サッカー映像の内容記述（映像内容の説明文生成）を行えるのは明らかである。
【００６１】
図５および図６に示した重要度値のみに基づき重要シーンを抽出する場合に、問題となるのは、重要度値の大きなものから優先的に選択するようにしても、重要度の値が同じであるシーンが多く、どこから抽出すればよいかわからない、換言すれば、全体としては適切に映像シーンを抽出しているものの、最も適切な映像シーンが抽出されるか否かは定かでないという不具合がある。
【００６２】
そこで本実施の形態では、その際に上記の試合の流れ情報を考慮し、攻勢になっているところ（すなわちＡチームの中盤以降）から優先してシーンを選択していく。このときのシーン抽出の割合は以下のようになっている。
激しい攻防→ Ｈチーム１、Ａチーム１
攻勢 → 攻勢チーム１
攻めあぐね→ Ｈチーム0.25、Ａチーム0.25
【００６３】
上記により、今回の結果におけるシーンの抽出比率は、表１の通りとなる。本実施の形態では、この抽出比率に合うように比例代表方式で、シーンの抽出を行なっていくので、試合の流れを考慮した映像シーンの抽出を行って、さらに適切なダイジェスト映像の作成を行うことができる。
【００６４】
前述したように本実施の形態では、サッカーの試合を複数の時間帯（前半を序盤、中盤、終盤の３つに、同様に後半を序盤、中盤、終盤の３つの区間に分ける）に分割し、それぞれの時間帯において各チームが「攻勢」であるか、「劣勢」であるか、「攻めあぐね」ているかを判定し、この各時間帯の判定結果を試合の流れ情報として生成する。換言すれば、各時間帯の判定結果の変化（流れ）を試合の流れとして代用する。このようにして生成した試合の流れ情報および重要度を用いて映像シーンの抽出を行うことにより、試合の流れを考慮したダイジェスト映像が作成される。
【００６５】
【発明の効果】
以上説明したように、本発明のサッカー映像ダイジェスト作成方法（請求項１）によれば、サッカーの試合を複数の時間帯に分割し、それぞれの時間帯において各チームが「攻勢」であるか、「劣勢」であるか、「攻めあぐね」ているかを判定して試合の流れ情報を生成し、試合の流れ情報および重要度を用いて映像シーンの抽出を行うため、試合の流れを考慮した映像シーンの抽出を行って、さらに適切なダイジェスト映像の作成を行うことができる。
【００６６】
また、本発明のサッカー映像ダイジェスト作成方法によれば、時間帯のそれぞれにおいて、各時間帯内のボール所持レベルを積分し、その値がある一定値以上ならば「攻勢」、一定値以下であるならば「劣勢」、それ以外であれば「攻めあぐね」と判定するため、もっともらしい（確からしい）試合の流れ情報を容易に生成することができる。
【００６７】
また、本発明のサッカー映像ダイジェスト作成方法によれば、「攻めあぐね」と判定された場合、その時間帯内にシュート数が一定数以上あったか否か判定し、一定数以上あった場合には、「攻めあぐね」の判定を「攻勢」に変更するため、実際の試合の流れにより近い試合の流れ情報を生成できる。
【００６８】
また、本発明のサッカー映像ダイジェスト作成方法によれば、試合の流れ情報は、それぞれの時間帯に対して判定された各チームの判定結果の組み合わせによって、ホームチームが攻勢でアウェイチームも攻勢の場合に「激しい攻防」が生成され、ホームチームが攻勢でアウェイチームが劣勢の場合に「ホームチーム攻勢」が生成され、ホームチームが劣勢でアウェイチームが攻勢の場合に「アウェイチーム攻勢」が生成され、ホームチームが劣勢でアウェイチームも劣勢の場合に「両者攻めあぐねる」が生成されるため、ダイジェスト映像の抽出に有効に利用可能な試合の流れ情報を生成することができる。
【００６９】
また、本発明のサッカー映像ダイジェスト作成方法によれば、映像シーンを抽出する場合、最初に重要度の高い順に選択し、同じ重要度のシーンが複数あった場合には、各時間帯の試合の流れ情報をチェックし、「激しい攻防」、「攻勢」、「攻めあぐね」の３つの状態に対して、それぞれ重みをつけてシーン抽出の比率を設定することにより抽出する映像シーンの優先度を算出し、優先度の高い順に選択するため、試合の流れを考慮した映像シーンの抽出を行って、さらに適切なダイジェスト映像の作成を行うことができる。
【図面の簡単な説明】
【図１】本実施の形態のサッカー映像ダイジェスト作成方法を適用したダイジェスト作成装置の概略構成図である。
【図２】本実施の形態のエリア情報を示す説明図である。
【図３】構造インデックスおよび動作インデックスの属性リストの例を示す説明図である。
【図４】サッカー映像にＵＮＤＯインデックスを含めた映像インデックスを付加した例を示す説明図である。
【図５】本実施の形態に基づく実験データ（実施例）のホームチームの重要度値（前半）を示す説明図である。
【図６】本実施の形態に基づく実験データ（実施例）のアウェイチームの重要度値（前半）を示す説明図である。
【図７】本実施の形態におけるホームチームのボール保持レベルパラメータの計算例を示す説明図である。
【図８】本実施の形態におけるアウェイチームのボール保持レベルパラメータの計算例を示す説明図である。
【符号の説明】
１００ ダイジェスト作成装置
１０７ ダイジェスト作成ソフト[0001]
BACKGROUND OF THE INVENTION
The present invention, for soccer video, for each home team and away team, the time that the player holds the ball and the ball possession level indicating the area information and the ball movement level indicating the distance of movement of the ball per unit time As a situation level, and using the calculated situation level to determine the importance of a video scene in a soccer video and create a digest video, more specifically, digest video extraction The present invention relates to a soccer video digest creation method that allows a more appropriate digest video to be created by applying the concept of a game flow to the game.
[0002]
[Prior art]
In recent years, digitalization of broadcasting has rapidly progressed on a global scale, and preparations for BS (Broadcast Satellite) digital broadcasting and terrestrial digital broadcasting are steadily progressing. As a result, the viewing mode of the television changes rapidly, and not only conventional real-time viewing but also storage-type viewing and non-linear viewing modes are possible.
[0003]
Here, the digest creation system in the non-linear viewing mode proposed by the present applicants will be described. First, the applicant searches for a video scene that is assumed to be an important scene using the video index, and creates a digest version of the video (digest video). A digest creation system was devised. In addition, when a digest video is created using a video index, a digest creation device that creates a digest video reflecting the preference of a viewer (user) who uses the video has also been proposed. Furthermore, there is also a function for adding natural expressions to the video search results by adding emotional expressions depending on the viewer's preference, and by adding connection expressions to fragmented explanations for each video scene. is suggesting.
[0004]
By using this technology, a digest video of a semantically important video scene can be created from the video with the fragmentary information describing the content of the video scene added as a video index using the video index. Can do.
[0005]
However, in this case, the target video is a video that is relatively structurally clear as a whole, and the video that is structurally ambiguous as a whole must always create a digest video accurately. There was a problem that it was not possible. Here, the structurally clear video means that the conditions for structuring the video such as innings and at-bats are ruled in advance as in baseball video, and the progress of the game itself (video flow) itself Is an image that can be easily structured, and an index (structure index) that represents the structure can be mechanically added. A structurally ambiguous image is a condition for structuring an image like a soccer image. Since there are few and the switching of offense and defense is not clear, it is difficult to determine the structure, and it is difficult to mechanically add an index representing the structure (structure index).
[0006]
Therefore, the present applicants have proposed a soccer video digest creation method that creates a digest video by extracting a video scene having a high degree of importance from a structurally ambiguous sports video such as a soccer video.
[0007]
[Problems to be solved by the invention]
However, in the method of creating a digest video by extracting a highly important video scene from a structurally ambiguous sports video proposed by the applicant, a shot scene or the like is obtained by extracting a video scene having a high importance. Although it was possible to extract important scenes to some extent, it was not possible to extract video scenes considering the flow of the game. In other words, it was not possible to create a digest video considering the flow of the game.
[0008]
Specifically, when creating a digest video of a soccer program, in addition to determining an important scene, it is important to represent the flow of a game up to that scene. For example, when there are multiple scenes that are determined as important scenes, it is a scene that has shot in a depressing situation, or a one-shot scene that is during a fierce attack that shoots many shots Depending on whether or not it is, the importance will naturally be different. Also, regarding the flow of such a game, “The home team was dominant at the beginning of the game, but the away team gradually attacked and took the first goal” or “The home team was always feeling However, I would like to see a description such as “The scene where the away team finally attacked and scored one point”. In the conventional methods, it was not possible to select a scene considering such a game flow.
[0009]
The present invention has been made in view of the above, and an object of the present invention is to extract a video scene in consideration of the flow of a game and create a more appropriate digest video.
[0010]
[Means for Solving the Problems]
In order to achieve the above object, the method for creating a soccer video digest according to claim 1 includes basic event information during a soccer game such as a game start, first half start, free kick, corner kick, shot, goal, possession of a ball, etc. The ball indicating the time that the player holds the ball and the area information for each home team and away team for the soccer video with the area information indicating the ball position at the time of the event as index information. Calculate the possession level and the ball movement level indicating the movement distance of the ball per unit time as the situation level, determine the importance of the video scene in the soccer video using the calculated situation level, Said In the soccer video digest creation method that extracts video scenes with high importance and creates them as digest videos, In each of the time zones in which the soccer game time is divided into a plurality of times, an integrated value of the ball possession level of each team is calculated, and each team is attacked from a comparison result between the calculated integrated value and a predetermined threshold value. Alternatively, it is determined whether the team is inferior, and the combination of the determination results of the teams in the respective time zones, the flow information of the game for each time zone, when the home team is offensive and the away team is also offensive Generated as `` violent offense and defense '', generated as `` home team offensive '' when home team is offensive and away team is inferior, generated as `` away team offensive '' when home team is inferior and away team is offensive However, when the home team is inferior and the away team is also inferior, it is generated as “both attacking together” and the flow of the generated game is According to the information and the importance, the video scenes are extracted in the order of the importance, and when there are a plurality of the video scenes having the same importance, the flow information of the game in each time zone is checked, The priority of the video scene to be extracted is calculated by assigning a weight to each of the three states of “violent offense and defense”, “attack”, and “attacking”, and setting the ratio of scene extraction, Extract the video scenes in descending order of priority It is characterized by that.
[0011]
According to the present invention, a soccer game is divided into a plurality of time zones, for example, an early stage, a middle stage, and an end stage, and each team is “offensive” or “inferior” in each time zone. , Judge whether you are attacking. The determination result of each time zone is generated as game flow information. In other words, the change (flow) of the determination result in each time zone is used as the game flow. By extracting video scenes using the game flow information and importance generated in this way, a digest video considering the game flow is created.
[0012]
A soccer video digest creation method according to claim 2 is the soccer video digest creation method according to claim 1, If it is determined that the “attacking”, it is determined whether or not the number of shots within a certain time period is greater than a certain number, and if there is a certain number or more, the determination of “attacking” is changed to the “attack”. change It is characterized by that.
[0013]
Moreover, the soccer video digest creation method according to claim 3 1 or 2 In the soccer video digest creation method described in Priority is given to selecting the video scene from where it is “offensive” It is characterized by that.
[0016]
DETAILED DESCRIPTION OF THE INVENTION
Hereinafter, an example in which the soccer video digest creation method of the present invention is applied to a digest creation device will be described in detail with reference to the accompanying drawings.
[0017]
(This embodiment)
About the digest creation device to which the soccer video digest creation method of the present embodiment is applied,
(1) About soccer footage as a structurally ambiguous sports footage
(2) Configuration of digest creation device
(3) Basic video index used in this embodiment
(4) Play section
(5) UNDO index
(6) Ball holding level parameter and ball movement level parameter
(7) Judgment of important scenes
(8) Examples
This will be explained in the order.
[0018]
(1) About soccer video as structurally ambiguous sports video,
In the present invention, a soccer video that is structurally ambiguous is used as a target for creating a digest video. It is difficult to create a digest of a soccer video compared to a baseball video that is structurally clear. The reason is due to the following characteristics of soccer video.
(1) The structure is not clear. Therefore, it is difficult to add an index indicating the structure.
(2) It is difficult to calculate the importance level of the video scene because the numerical information indicating the type of the index and the content of the game is small (hereinafter, an index indicating the importance level of the video scene is called an importance level determination parameter).
(3) The movement is intense and there are many cooperative movements. Therefore, it is difficult to follow the development of the game with the camera, and there are many large-sized videos, making it difficult to grasp the entire play and to determine the contents.
[0019]
The characteristics and problems of such a soccer video (structurally ambiguous sports video) will be described.
First of all, the feature (1) is not clear. The baseball image can be mechanically added with an index indicating the structure, such as innings and at-bats, and can be automatically structured. This means that it is difficult to structure. As a result, in the case of a soccer video, the index addition indicating the structure tends to be subjective. In addition, since it is difficult to determine the structure, there are many reworks such as adding an index while returning the video and confirming it again.
[0020]
Regarding the calculation of the importance determination parameter (2), the content of the game can be imagined from the score book when compared with the baseball video. In the case of baseball video using the technology already proposed by the present applicant, there are a lot of numerical information that can be expressed in a certain range of numerical values such as score information, SBO (Strike Ball Out) information, progress information, etc. The importance determination parameter can be calculated by combining them with the video index information mechanically added. For example, when an attack level representing an aggressive importance is defined as one of the importance determination parameters, the attack level is based on an aggressive index such as hit, home run, outing, and score, and score information. It is possible to easily define a rule for calculating an attack level (importance level determination parameter).
[0021]
On the other hand, in the case of a soccer video, since there is little numerical information that is an objective index and there are few types of indexes, it is difficult to calculate an importance determination parameter. For example, what can be mechanically added as an aggressive index is a shot or goal, and what can be used as objective numerical information is about score information and time information. This alone simplifies the rule for calculating the attack level (importance determination parameter), creates a digest video in which only the shooting scene is cut out, and cannot accurately represent the content of the game.
[0022]
(3) is due to the nature of soccer itself, but soccer video has cooperative pass play, so it is more difficult to grasp the role of the player than baseball video. In particular, because there are many large-scale videos in soccer video, or because other players' play is shown, it is difficult to always identify players, and other parts of cooperative play cannot be seen. There are also difficulties. Furthermore, even if you can see the movements of all the players nearby, in order to determine the importance of the play and add an index, identification ability is required, and the evaluation of the content of the play is determined Since the degree of dependence on the person increases, it is difficult to make a uniform determination (difficulty in content determination).
[0023]
For this reason, it is difficult to create an index system and add an index to a soccer video (a sports video that is structurally ambiguous).
For example, consider a case where a semantically grouped flow such as a soccer attack scene is cut out as a digest video. In soccer, the path of attack is very dynamic, such as frequent cuts in the path. Since the index producer cannot read the future, many operations such as cancellation of index addition, extension of period, and re-attachment occur on the way. This is even more difficult when delivering in real time. In some cases, the index distribution is performed on the soccer video in real time. However, since the index creator may perform UNDO (invalidation) later, it is necessary to immediately add and distribute the index.
[0024]
In the case of a baseball video, it is possible to calculate the importance determination parameter only with fragmented index information and structure information indicating a certain point on the video. On the other hand, an importance determination parameter cannot be calculated for a soccer video whose structure is not clear and which is developed quickly due to a lack of index information and a lack of structure information.
[0025]
Therefore, in the present embodiment, when determining the importance level of a video scene in a soccer video (structurally ambiguous sports video), an importance level determination parameter is calculated based only on fragmented index information and structural information. Rather than obtaining an auxiliary parameter (situation level parameter) representing the situation on the video from the fragmented index information, the importance level determination parameter is calculated by combining the situation level parameter and the fragmented index information. Create a digest video.
[0026]
In addition, when creating a digest video of a soccer program, it is important to express the flow of a game up to that scene in addition to determining an important scene. For example, when there are multiple scenes that are determined as important scenes, it is a scene that has shot in a depressing situation, or a one-shot scene that is during a fierce attack that shoots many shots Depending on whether or not it is, the importance will naturally be different. Also, regarding the flow of such a game, “The home team was dominant at the beginning of the game, but the away team gradually attacked and took the first goal” or “The home team was always feeling However, I would like to see a description such as “The scene where the away team finally attacked and scored one point”. Therefore, in this embodiment, there is provided a soccer video digest creation method capable of extracting a video scene in consideration of such a game flow and creating a more appropriate digest video.
[0027]
(2) Configuration of digest creation device
FIG. 1 shows a schematic configuration of a digest creation device according to the present embodiment, FIG. 1A shows an overview of the digest creation device 100, and FIG. 1B shows a hardware configuration. As a hardware configuration of the digest creation device 100, a CPU 101 that executes processing based on software described later, a ROM 102 that stores a boot program, a RAM 103 that is used as a work area of the CPU 101, and various input operations are performed. Keyboard 105 and mouse 106, a display device 104 such as a CRT or a liquid crystal display having a display screen for displaying a soccer video, and a digest creation software 107 for realizing the soccer video digest creation method of the present embodiment. At least a hard disk 108 that stores various application programs and data, an interface unit 109 such as various drivers and network adapters for connecting to external devices, and a bus 110 that connects the above units. There.
[0028]
(3) Basic video index used in this embodiment
In the present embodiment, as a basic video index (event information) added to a soccer video, a structure index and an action index are provided.
[0029]
(1) Structure index
The structure index is an index (event information) indicating the structure of the game such as the start of the game, the end of the game, or the end of the first half / second half. However, since there are fewer types than structural sports videos (baseball videos), detailed structuring is not possible only with a structural index.
[0030]
(2) Motion index
This is an index (event information) indicating an action (in other words, an attack action) of a player holding the ball in a soccer video, and is classified into the following three indexes.
1) Separation index: This is an action index for creating a segment of a video scene in a play called set play, such as throw-in, free kick, corner kick, goal kick, fouling and the like. In other words, it is an index indicating an operation that becomes a break of the video scene in the operation of the target sport.
[0031]
2) Ball possession index: This is an index (event information) indicating who is holding the ball in which area of the ground due to operations such as passing, receiving, and cutting. It is added separately for each home team and away team. Occurs when a player who has the ball is switched, or when an action such as taking a pass is changed from receiving a pass.
In general, since the time from pass acceptance to pass pass is very short, it is assumed that the pass pass operation and pass pass operation are usually indicated only by the pass pass operation. In the present embodiment, it is determined that the dribbling operation occurs during the time from pass receipt to pass issue. That is, this ball possession index is information indicating the action of each team or player who is performing an attack action in the action of the target sport.
[0032]
3) Shoot and goal index: The index generated when the ball is kicked toward the goal is the shot index, and if the shot is successful, it becomes the goal index.
[0033]
4) UNDO index: This is an index for making UNDO (invalid) the other operation index added immediately before.
[0034]
Each operation index is set with the time code at which the event occurred. In addition to the team name and the player name, information such as the current ball position (area information to be described later) is set as an attribute (attribute information) in the motion index as necessary.
[0035]
Here, the area information will be described. Each action index is defined by dividing the ground (area) where the soccer game is played into multiple areas as information indicating the place where the corresponding action was performed (corresponding to the ball position in the soccer video) Area information is set.
[0036]
As shown in FIG. 2, the area information (ball position) of the present embodiment is roughly divided into seven areas for the opponent team (home team (H), away team (A)). The symbol (area name) was set as area information. The area names of HG, H1, H2, C, A1, A2, and AG shown in FIG. 2 are reversed left and right when the defensive position changes between the first half and the second half. A numerical value set for each area represents a weighting factor (AP: used for calculating importance) of the corresponding area.
[0037]
FIG. 3 shows a state in which a structure index and an action index are added to a certain soccer program. A team name, a player name, an area (area name), and the like are set in the attribute list of the structure index and the motion index. Although omitted in FIG. 3, time codes are set as attributes for all indexes. The ball holding index is indicated by the player name for convenience.
[0038]
(4) Play section
The section delimited by the delimiter index in the action index is defined as the play section. In a soccer video, the segment is large and insufficient only by segment division based on the structure index. Therefore, when determining the importance of the video scene on a segment basis, the importance is calculated using the play segment. In addition to the break index, a play section may be defined using a break index, a shoot, and a goal index.
[0039]
Also, each play section is divided into scenes called attack sections, with the offense and defense being changed by a cut operation in the ball possession index. The attack section indicates a scene in which each team has a ball.
[0040]
(5) UNDO index
Next, the role of the UNDO index will be described. The index producer attaches an action index in real time to a game flow that cannot be seen in advance, but it is expected that the action index may need to be corrected due to a change in the game flow.
For example, in a situation where everyone thinks that “the forward player has shot”, it is often the case that a pass was sent out instead of a shot. In such a case, if an UNDO index is added and distributed, the receiving terminal device side can interpret all video indexes including the UNDO index and finally obtain only valid video indexes as index information. . On the index producer side, if there is a time allowance for distribution, the operation index specified by the UNDO index can be deleted or corrected, and the corrected video index can be distributed.
[0041]
FIG. 4 is an explanatory diagram showing an example in which a video index including an UNDO index is added to a soccer video. The soccer video is structured into two scenes of the first half and the second half by the structure index 401. Further, a play section 403 is defined by a break index 402. In FIG. 4, it is thought that Nagai has shot, and after adding a shoot index, since it was a pass, an UNDO index 404 is added and corrected.
[0042]
As shown in FIG. 4, most of the motion index names are event names, whereas only the ball possession index has the player name as the index name. The reason for this is that, when performing index addition work in real time, the ball possession index is often added continuously, and the use of the player name makes the work more efficient.
[0043]
(6) Ball holding level parameter and ball movement level parameter
Next, the ball holding level parameter (Ps) and the ball movement level parameter (Pb) are defined as follows based on the basic video index (particularly the motion index) described above. The ball holding level parameter indicates the time that the player holds the ball and its area, and the ball movement level parameter indicates the moving distance of the ball per unit time.
[0044]

[0045]

[0046]
(7) Judgment of important scenes
Next, on the basis of the video index (event information), the following importance level determination parameters are defined in order to determine an important scene (a video scene with a high level of importance) of a soccer program.
[0047]

[0048]
The importance value is calculated by adding the weight value of the ball holding level parameter (Ps) and the ball movement level parameter (Pb) to the attack level value.
[0049]
(8) Examples
Based on the configuration and operation described above, a digest video was created from the soccer video. The experimental data shown in Figs. 5 and 6 show that (attack level + ball holding level parameter) is the importance value for each team in the J-League game "Verdi Kawasaki vs. Gamba Osaka" on March 12, 2000. This is a calculated example. In the figure, the vertical axis is the importance value obtained by calculation, and the horizontal axis is the time axis of the game progress.
[0050]
By the way, it is possible to determine an important scene such as a shoot scene to some extent from the above graph of importance values. However, in the case of generating a digest of a soccer program, in addition to determining an important scene, it is important to express the flow of a game up to that scene.
[0051]
For example, if there are multiple scenes that are determined as important scenes, it is a scene that has shot in a depressing situation, or a one-shot scene that is being attacked by a number of shots Depending on whether or not it is, the importance will naturally be different. In addition, regarding the flow of such a game, “The home team was dominant at the beginning of the game, but the away team gradually attacked and took the first goal” or “ However, I would like the explanation to be something like "a scene where the away team finally attacked and scored one point".
[0052]
Therefore, in the present invention, video scenes are extracted (selected) in consideration of the game flow as follows.
[0053]
First, in order to express the flow of this game, the following situation levels are calculated. This situation level is calculated for each of the home and away teams.
[0054]
(1) Divide the first half and the second half into several sections (time zones).
In the case of a soccer game, it is usually divided into three sections of about 15 minutes each of the beginning, middle and end.
(2) The integrated value (situation level) of the ball holding level parameter is calculated for the divided section.
(3) If the integral value is greater than or equal to a certain value (> = α), it is determined that the team is offensive in that section.
(4) If the integrated value is less than a certain value (> = β), it is determined that the team is inferior in that section.
As an example, the values of α and β are each set to −10.
(5) The respective sections of the home team (H) and the away team (A) are determined.
(6) The final game flow information is generated by combining the results.
The policy at that time is as follows.
Team H is offensive and Team A is also offensive →
When the home team is offensive and the away team is also offensive, “violent offense and defense” is generated as game flow information.
H team is offensive and A team is inferior → H offensive
When the home team is offensive and the away team is inferior, “home team offensive (H offensive)” is generated as game flow information.
Team H is inferior and Team A is offensive → Attack of A
When the home team is inferior and the away team is offensive, “Away team offensive (A offense)” is generated as the game flow information.
Team H is inferior and Team A is also inferior →
When the home team is inferior and the away team is also inferior, “both attacks” is generated as game flow information.
However, even if it is determined to be “Agone Agune”, it is determined whether or not the number of shots has exceeded a certain number within that time period. change.
[0055]
Therefore, the game flow information is generated by the combination of the judgment results of each team determined for each time zone, and when the home team is offensive and the away team is offensive, “violent offense and defense” is generated. A home team offensive is generated when the away team is inferior due to an offense, an “away team offensive” is generated when the home team is inferior and the away team is offensive, the home team is inferior and the away team is also inferior Will generate a “both attacks”.
[0056]
When each section is determined as described above, it is desirable that the extracted important scene reflects the above determination in the scene extraction for digest creation. Therefore, scene extraction is performed according to the following extraction rule.
[0057]
First, basically, selection is made in descending order of importance.
Secondly, when there are a plurality of scenes having the same importance level, the scene selection is made based on the predetermined scene extraction ratios for the three states of intense attack and defense, offensive, and offensive.
For example, the ratio of scene extraction is determined as follows for three states of intense offense and defense, offense, and offense.
Fierce offense and defense → H team 1, A team 1
Offensive → Offensive Team 1
Attack Agune → H team 0.25, A team 0.25
[0058]
The result of calculating the above-described situation level and applying the extraction rule to the game shown in FIGS. 5 and 6 is shown below.
First, the situation level is calculated. For this purpose, it is necessary to integrate the value of the ball holding level parameter for each section. 7 and 8 show the calculation results of the ball holding level parameter of each team. Table 1 shows values obtained by integrating the above values with respect to three sections (early, middle, and final) and their determination results.
[0059]
[Table 1]

[0060]
From the above, the first half of this soccer game
Early: Both home team and away team were attacking,
Midfield: Away team turned to offensive,
End: The away team continues to be offensive.
It can be expressed that the game flow. Also, since the flow of the game becomes clear in this way, it is clear that the content description of the soccer video (description of the video content) can be performed using the flow information of the game.
[0061]
When extracting an important scene based only on the importance values shown in FIG. 5 and FIG. 6, the problem is that even if priority is selected from those having a large importance value, the importance value is There are many scenes that are the same, and I do not know where to extract them.In other words, although the video scene is properly extracted as a whole, it is not certain whether the most appropriate video scene is extracted. There is.
[0062]
Therefore, in the present embodiment, in consideration of the flow information of the game at that time, the scene is selected with priority from the place where it is offensive (that is, the middle part of the A team). The ratio of scene extraction at this time is as follows.
Fierce offense and defense → H team 1, A team 1
Offensive → Offensive Team 1
Attack Agune → H team 0.25, A team 0.25
[0063]
As described above, the scene extraction ratio in this result is as shown in Table 1. In the present embodiment, scenes are extracted by a proportional representative method so as to match this extraction ratio. Therefore, video scenes are extracted in consideration of the flow of the game, and a more appropriate digest video is created. be able to.
[0064]
As described above, in this embodiment, the soccer game is divided into a plurality of time zones (the first half is divided into three sections of the early, middle and final stages, and the second half is divided into three sections of the early, middle and final stages). In each time zone, it is determined whether each team is “offensive”, “inferior”, or “offensive”, and the determination result of each time zone is generated as game flow information. In other words, the change (flow) of the determination result in each time zone is used as the game flow. By extracting video scenes using the game flow information and importance generated in this way, a digest video considering the game flow is created.
[0065]
【The invention's effect】
As described above, according to the soccer video digest creating method (claim 1) of the present invention, a soccer game is divided into a plurality of time zones, and each team is “offensive” in each time zone, Video that considers the flow of the game, because it determines whether it is “inferior” or “attacking” and generates the game flow information, and extracts the video scene using the game flow information and importance It is possible to create a more appropriate digest video by extracting a scene.
[0066]
Also, how to create a soccer video digest of the present invention To the law According to this, in each time zone, the ball possession level in each time zone is integrated, and if the value is above a certain value, “offensive”, if it is below a certain value, “inferior”, otherwise Since it is determined to be “attacking,” plausible (probable) game flow information can be easily generated.
[0067]
Also, how to create a soccer video digest of the present invention To the law According to this, if it is determined to be “Aggression”, it is determined whether or not the number of shots has exceeded a certain number within that time period. Since the change is made, it is possible to generate game flow information closer to the actual game flow.
[0068]
Also, how to create a soccer video digest of the present invention To the law According to the game flow information, “strong attack and defense” is generated when the home team is offensive and the away team is also offensive by the combination of the judgment results of each team judged for each time zone. `` Home team offensive '' is generated when the away team is inferior and the away team is inferior, `` Away team offensive '' is generated when the home team is inferior and the away team is offensive, the home team is inferior and the away team is also inferior In this case, since “both attacks” is generated, it is possible to generate game flow information that can be effectively used for extraction of the digest video.
[0069]
Also, how to create a soccer video digest of the present invention To the law According to the above, when extracting video scenes, first select them in order of importance, and if there are multiple scenes with the same importance, check the flow information of the game in each time zone, In order to calculate the priority of the video scene to be extracted by setting the ratio of scene extraction by weighting each of the three states of “offensive” and “offensive”, and selecting them in order of priority. It is possible to extract a video scene considering the flow of the game and create a more appropriate digest video.
[Brief description of the drawings]
FIG. 1 is a schematic configuration diagram of a digest creation device to which a soccer video digest creation method of the present embodiment is applied.
FIG. 2 is an explanatory diagram showing area information according to the present embodiment.
FIG. 3 is an explanatory diagram illustrating an example of an attribute list of a structure index and an operation index.
FIG. 4 is an explanatory diagram showing an example in which a video index including an UNDO index is added to a soccer video.
FIG. 5 is an explanatory diagram showing importance values (first half) of home teams of experimental data (examples) based on the present embodiment.
FIG. 6 is an explanatory diagram showing importance values (first half) of away teams of experimental data (examples) based on the present embodiment.
FIG. 7 is an explanatory diagram showing a calculation example of a ball holding level parameter of a home team in the present embodiment.
FIG. 8 is an explanatory diagram showing a calculation example of a ball holding level parameter of an away team in the present embodiment.
[Explanation of symbols]
100 digest creation device
107 digest creation software

Claims (3)

Soccer with index information that includes basic event information during soccer matches such as match start, first half start, free kick, corner kick, shoot, goal, possession of ball, and area information that indicates the position of the ball when that event occurs For the video, for each home team and away team, the time the player holds the ball and the ball possession level indicating the area information and the ball movement level indicating the ball movement distance per unit time are used as the situation level. calculated, using the calculated status levels to determine the importance of the video scene in a soccer video, the soccer video digest generation method for generating the a high importance digest extracted video scene image,
In each of the time zones in which the soccer game time is divided into a plurality of times, the integrated value of the ball possession level of each team is calculated,
Determine whether each team is offensive or inferior from the comparison result of the calculated integral value and a predetermined threshold,
According to the combination of the determination results of each team in each time zone, the flow information of the game for each time zone is generated as “violent offense and defense” when the home team is offensive and the away team is also offensive, If the team is offensive and the away team is inferior, it will be generated as `` Home Team Offensive '', if the home team is inferior and the away team is offensive, it will be generated as `` Away Team Offensive '' and the home team will be inferior and the away team If it is inferior, it will be generated as “both attacks”
The video scenes are extracted in descending order of importance based on the generated flow information and importance of the game, and when there are a plurality of video scenes having the same importance, The flow information is checked by setting the ratio of scene extraction by weighting each of the three states of “violent offense and defense”, “offensive”, and “attacking”. A soccer video digest creation method , wherein priority is calculated and the video scenes are extracted in descending order of priority . If it is determined that the "attack Agne", it is determined whether the number of shoots within that period of time was more than a certain number, when a predetermined number or more, the determination of the "attack Agne" to "the offensive" The soccer video digest creation method according to claim 1, wherein the method is changed. 3. The soccer video digest creation method according to claim 1 or 2, wherein the video scene is selected with priority from the place of the "offensive".

Images