JP2007335984A - Event end point extraction apparatus, event end point extraction method, and extraction program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide an event end point extraction apparatus for accurately extracting an event end point of a sports image being an object of digest viewing. <P>SOLUTION: The event end point extraction apparatus 100 includes: a sports image storage section 101 for storing the sports image whose event start time is determined; a cut-point detection section 102 for detecting cut-points after the event start time from the sports image in the sports image storage section 101; a cut length calculation section 104 for classifying a type of a cut length between the cut-points detected by the cut-point detection section 102 depending on the short/long cut length; an appearance pattern storage section 112 for storing an appearance pattern of the type of the cut-length in cross reference with the event end point; an event end point extraction section 111 for extracting the event end point corresponding to the appearance pattern of the classified type by referencing the appearance pattern storage section 112; and an output section 113 for outputting the event end point extracted by the event end point extraction section 111 and the event start point. <P>COPYRIGHT: (C)2008,JPO&INPIT

Description

  The present invention relates to an event end point extraction device, and more particularly to an event end point extraction device that extracts an event end point of a sports video to be viewed.

In recent years, in order to perform digest viewing of a video such as a baseball game, a specific scene included in the video may be extracted. Under such circumstances, the conventional apparatus extracts a specific scene from the length of the cut length of the video. Specifically, the conventional apparatus first classifies a section with a short cut length as a video scene with an action, and classifies a section with a long cut length as a gentle video scene. Then, this device accurately extracts a video scene with the above action in consideration of the intensity of motion of the video (for example, Patent Document 1).
In another conventional apparatus, a cut having a total cut length composed of a plurality of cuts of video is 15 seconds or 30 seconds is extracted. If there is a silent section at the beginning and the end of the cut, this apparatus extracts the section as a CM scene (for example, Patent Document 2).

JP 2006-14084 A JP-A-8-339379

However, in the apparatuses described in Patent Document 1 and Patent Document 2, it is not possible to extract a series of scenes including long and short cut lengths as one unitary scene. There is a disadvantage that the event end point included in the sports image cannot be accurately extracted.
Therefore, the present invention has been made to solve the above-described inconveniences, and an object thereof is to accurately extract an event end point of a sports video that is a subject of digest viewing.

  To achieve the above object, the present invention detects a sports video storage unit that stores a sports video for which an event start point has been determined, and detects cut points after the event start point from the sports video in the sports video storage unit The cut point detection unit, the cut length classification unit for classifying the type of cut length between the cut points detected by the cut point detection unit according to the short length of the cut length, and the appearance of the type of cut length An appearance pattern storage unit that stores a pattern in association with an event end point; an event end point extraction unit that extracts an event end point corresponding to an appearance pattern of the classified type with reference to the appearance pattern storage unit; An output unit that outputs the event end point extracted by the event end point extraction unit in association with the event start point.

  ADVANTAGE OF THE INVENTION According to this invention, the event end point of the sports image | video used as the object of digest viewing can be extracted correctly.

Hereinafter, embodiments of the present invention will be described with reference to FIGS.
FIG. 1 is a block diagram showing a configuration example of an event end point extraction device according to an embodiment of the present invention.
In FIG. 1, an event end point extraction device (hereinafter referred to as “extraction device”) 100 includes a sports video storage unit 101, a cut point detection unit 102, a cut point storage unit 103, a cut length classification unit 104, and a classification reference storage unit 105. Have. Further, the extraction device 100 includes a telop detection unit 106, a telop storage unit 107, a first classification correction unit (classification correction unit) 108, a second classification correction unit (classification correction unit) 109, and a third classification correction unit. (Classification correction unit) 110 is included.
Further, the extraction device 100 includes an event end time extraction unit (event end point extraction unit) 111, an appearance pattern storage unit 112, and an output unit 113. In the present embodiment, a case where a server device is used as the extraction device 100 will be described as an example, but a computer such as a personal computer may be applied.

The sports video storage unit 101 stores sports videos that are subject to digest viewing. Specifically, as illustrated in FIG. 2, the sports video storage unit 101 stores a sports video (moving image) in units of frames configured with a time axis t. The sports video includes various telops depending on the video scene.
In the present embodiment, the sports video is, for example, a video of a professional baseball broadcast, and the telop included in this video includes types such as SBO, progress status, battle score, inning (for example, twice table), etc. There is. SBO is an abbreviation for strike count (S), ball count (B), and outcount (O).
FIG. 2 illustrates a situation in which a sports video d1 of a certain frame includes SBO telop T1, progress status telop T2, score telop T3, and inning telop T4. Note that the telop means display, and the form is not limited.

Further, in the sports video of FIG. 2, an event start time (event start point) IN is determined in advance. The event start time IN means the start time of the video scene that is the subject of the digest viewing, that is, the position (time) of the frame at which the video scene starts. For example, as the event start time IN, the position of a frame representing a video scene at which pitching is started is assumed.
As the event start time IN, for example, an elapsed time (time) from the first frame of the sports video is used, but the form is not limited as long as the position can be specified. For example, the counter value of the reproduction counter may be used as the event disclosure time IN (the same applies to the cut point and event end time described later).

In FIG. 2, the first cut point, the second cut point, and the third cut point are illustrated as cut points that appear after a predetermined time has elapsed from the event start time IN. Furthermore, a first cut length, a second cut length, and a third cut length are shown. The first cut length represents a section between the event start time IN and the first cut length, and the second cut length represents a section between the first cut point and the second cut point. The third cut length represents a section between the second cut point and the third cut point.
The first cut length is shown as a medium time, and the second cut length and the third cut length are both shown as a short time. These points will be described later.

Returning to FIG. 1, the cut point detection unit 102 reads a sports video from the sports video storage unit 101, detects cut points after the event start time IN, and registers them in the cut point storage unit 103. The cut point means the transition of the video scene. For example, when the relay from one camera is switched to another camera, the position of the frame when it is relayed to another camera and first taken corresponds to the cut point.
In the present embodiment, as a method for detecting the cut point described above, for example, a method of detecting a boundary between successive frame sets photographed from one camera is used (see Patent Document 1). The known detection method may be applied.
By configuring in this way, the cut point detection unit 102 sequentially sets the position of the frame when a predetermined time has elapsed from the event start time IN, as shown in FIG. It is possible to detect the second cut point and the third cut point.

  The cut length classification unit 104 refers to the classification standard in the classification standard storage unit 105 and classifies the type of cut length between consecutive cut points in the cut point storage unit 103. In the present embodiment, according to the short length of the cut length, “short (first type)”, “medium (second type)” and “long (third) However, the number of these types may be two or four or more.

Here, an example of the classification criteria in the classification criteria storage unit 105 is shown in FIG.
The classification standard in FIG. 3 is composed of items of classification and cut length. In the classification item, three types of “short”, “medium”, and “long” are designated, and in the cut length item, the corresponding cut length range is designated in time notation.

Specifically, for example, a cut length of 0 seconds or more and less than 2.0 seconds (plus or minus α) is specified for the “short” classification, and 2.0 seconds (plus or minus) for the “medium” classification. α) A cut length of not less than 4.0 seconds is specified. For this reason, when the cut length is in the range of 2.0 seconds plus or minus α (preliminarily designated time), the cut length belongs to both “short” and “medium”. In other words, this is based on the following reason.
That is, when the cut length has the time specified in advance, the type of the cut length is set to “short” or “medium” only by the time of the cut length regardless of the content of the video scene of the cut length. This is because it is difficult to classify it into one of "
Therefore, in the present invention, in the first classification correction unit 108, the second classification correction unit 109, or the third classification correction unit 110, which will be described later, the video scene having the cut length is taken into consideration by the display pattern of the telop described above. Above, the process which correct | amends the type of the cut length mentioned above (which belongs to both “short” and “medium”) to either type is performed. This makes it possible to accurately classify all types of cut lengths.

  It should be noted that the above-specified time (for example, 2.0 seconds plus or minus α, especially the value of α) is preferably determined in consideration of the type of sports video, the specific content of the video scene, and the like. .

  With this configuration, the cut length classifying unit 104 classifies the types of cut lengths between consecutive cut points in the cut point storage unit 103 in accordance with the classification criteria of FIG. 3 (see FIG. 1). As a result of this classification, when the cut length has a preset time (in the case of two types), the cut length classification unit 104 indicates that (each cut point and the cut length after the event start time IN). Are included) to the first classification correction unit 108 to be described later.

  Returning to FIG. 1, the telop detection unit 106 detects the above-described telop (see FIG. 2) from the sports video in the sports video storage unit 10 and registers it in the telop storage unit 107. The telop registered at this time is associated with the elapsed time from the first frame of the sports video, for example.

  When the cut length classified by the cut length classification unit 104 has a preset time, the first classification correction unit 108 performs the first cut based on a display pattern of a specific telop that suggests a three-way video scene. The classification type of the cut length that appears next to the length and the “medium” cut length that appears first from the event start time IN is corrected (for example, the type is corrected to a preset “short” type). Classify. The following two patterns are set in advance as display patterns here.

That is, the first display pattern is based on a combination of SBO telop (specific telop) T1 that shifts from two strikes to change, and inning telop (specific telop, for example, 2 times front → 2 times back) T4 that changes inning. Display pattern (display pattern of a specific telop).
The second display pattern includes an SBO telop (specific telop) T1 in which the outcount is increased by 1 from 2 strikes, a progress situation telop (specific telop) T2 in which there is no runner when the outcount is increased by 1, and It is a display pattern (display pattern of a specific telop) composed of a combination of score telops (specific telops) T3 whose score does not change.
The specific telop described above is read from the telop storage unit 107 by the first classification correction unit 108.

  The second classification correction unit 109 corrects the classification type of only the first cut length based on the display pattern of a specific telop that suggests a batter-change video scene (for example, to a preset “short” type). Correct) and classify the type. The display pattern here is as follows. That is, this is a display pattern of an SBO telop (specific telop) T1 whose strike count or out count is other than zero. The SBO telop T1 is read from the telop storage unit 107 by the second classification correction unit 109.

  The third classification correction unit 110 corrects the classification type of the third cut length (runner home-in video scene) based on a display pattern of a specific telop that suggests a video scene with an additional score (for example, preset). (Corrected to “short” type) to classify the type. The display pattern here is as follows. That is, it is a display pattern of a score telop (specific telop) T3 indicating that a score has been entered. The SBO telop T1 is read from the telop storage unit 107 by the third classification correction unit 110.

  The event end time extraction unit 111 refers to the appearance pattern in the appearance pattern storage unit 112, and the cut length classification unit 104, the first classification correction unit 108, the second classification correction unit 109, or the third classification correction unit 110. The event end time (event end point) corresponding to the appearance pattern of the type classified as (including the combination of these parts) is extracted.

Here, the appearance pattern of the appearance pattern storage unit 112 is shown in FIG.
The two appearance patterns in FIG. 4 are stored in the appearance pattern storage unit 112. As the appearance pattern shown in FIG. 4A, a pattern in which the type of “long” first appears from the event start time IN (not shown) is shown. The end point of the cut length belonging to the “long” type is associated with the event end time OUT. Therefore, when the appearance pattern of the classified type is the appearance pattern of FIG. 4A, the event end time extraction unit 111 sets the end point of the cut length belonging to the “long” type as the event end time OUT. Extract.
As an appearance pattern shown in FIG. 4B, a pattern in which the “medium” type appears continuously from the event start time IN (not shown) until the “long” type first appears. It is shown. Then, the end point of the cut length belonging to the first “medium” type is associated as the event end time OUT. Therefore, when the appearance pattern of the classified type becomes the appearance pattern of FIG. 4B, the event end time extraction unit 111 determines the end point of the cut length belonging to the first “medium” type as the event end time. Extracted as OUT.

  Returning to FIG. 1, the output unit 113 outputs the event end time OUT extracted by the event end time extraction unit 111 and the event start time IN to an external device (for example, a computer).

In the present embodiment, each of the storage units 101, 103, 105, 107, and 112 described above corresponds to a storage device such as a memory or a hard disk. In addition, each of the above-described units 102, 106, 108 to 111, 113 corresponds to a processing device such as a CPU.
The above-described units 102, 106, 108 to 111, 113 operate according to the extraction program stored in the above-described storage device. This extraction program may be read from a computer-readable medium such as a CD-ROM.

Next, the processing procedure of the extraction apparatus 100 will be described with reference to FIG.
FIG. 5 is a flowchart illustrating an example of a processing procedure of the extraction apparatus 100. Here, the sports video (see FIG. 2) that is the subject of the digest viewing will be described as being stored in the sports video storage unit 101.
First, the cut point extraction unit 102 reads a sports video (including the event start time IN) from the sports video storage unit 101, and detects cut points after the event start time IN (step S101). The cut point detected at this time is stored in the cut point storage unit 103 by the cut point detection unit 102.

  Subsequently, the telop detection unit 106 (extraction device 100) reads out the sports video from the sports video storage unit 101, and detects all the telops included in the sports video of each frame from the sports video (step S102). The telop detected at this time is stored in the telop storage unit 107 by the telop detection unit 106.

  Next, the cut length classifying unit 104 acquires the cut points after the event start time IN detected by the cut point detecting unit 102 from the cut point storage unit 103, and the classification standard in the classification standard storage unit 105 (see FIG. 3). Based on the above, the cut lengths between consecutive cut points are classified (step S103). Specifically, the cut length classifying unit 104 determines the type (“short”, “medium” corresponding to the time of the cut length between successive cut points based on the classification standard (see FIG. 3) of the classification standard storage unit 105. ”,“ Long ”).

Next, the first classification correction unit 108 (extraction apparatus 100) has a case in which the cut lengths classified by the cut length classification unit 104 have a predetermined time (for example, 2.0 plus or minus α in FIG. 3). Then, it is determined whether or not there is a possibility of strikeout in the cut length that appears next to the cut length, that is, the first cut length and the “medium” cut length that appears first from the event start time IN (step S104).
Specifically, the first classification correction unit 108 determines whether or not a specific telop display pattern (any one of the first to second display patterns described above) suggesting a strikeout video scene is in the cut length. Judging. In step S104, if there is no cut length having a predetermined time, the process may skip from step S103 to step S110.

  As a result of the determination, if it is determined that there is no specific telop display pattern that suggests a strikeout video scene and there is no possibility of a strikeout (No in step S104), the first classification correction unit 108 The type of the cut length to be determined is corrected and classified as “short” (step S105), and the process proceeds to step S106. As a result, the type of the cut length having a predetermined time (for example, the cut length between the first cut point and the second cut point) is corrected to “short”.

  On the other hand, if it is determined that there is a possibility of strikeout (Yes in step S104), the process proceeds to step S106. In the case where the process proceeds to step S106, the first classification correction unit 108, for example, sets each classification of cut points and cut lengths after the event start time IN (including types corrected) as the second classification correction. Passed to part 109.

In step S106, the second classification correction unit 109 (extraction device 100) determines that the cut length classified by the cut length classification unit 104 has a predetermined time (for example, 2.0 plus or minus α). It is determined whether there is a batter change in the cut length, that is, the first cut length.
Specifically, the second classification correction unit 109 cuts a specific telop display pattern (an SBO telop T1 display pattern with a strike count or outcount of 0) that suggests a batter-change video scene. Determine if you are in the head.

  As a result of the above determination, when it is determined that there is no display pattern of a specific telop that suggests a batter change video scene and there is no batter change (No in step S106), the second classification correction unit 109 The type of the cut length to be determined is corrected and classified as “short” (step S107), and the process proceeds to step S108. As a result, the type of the cut length having a predetermined time (for example, the cut length between the first cut point and the second cut point) is corrected to “short”.

  On the other hand, if it is determined that there is a change in batter (Yes in step S106), the process proceeds to step S108. In the case of proceeding to step S108, the second classification correction unit 109 includes, for example, each classification of cut points and cut lengths after the event start time IN (including those that have been subjected to type correction in steps S105 and S107). The data is passed to the third classification correction unit 110.

In step S108, the third classification correction unit 110 (extraction apparatus 100) determines that the cut length classified by the cut length classification unit 104 has a predetermined time (for example, 2.0 plus or minus α). It is determined whether there is a score added to the cut length, that is, the third cut length.
Specifically, the third classification correction unit 110 determines whether a specific telop display pattern (score telop T3 display pattern indicating that a score has been entered) suggesting a video scene to which a score has been added has the cut length. Judge whether.

  As a result of the above determination, when it is determined that there is no specific telop display pattern that suggests a video scene for which a score has been added and no score has been added (No in step S108), the third classification correcting unit 110 The type of the cut length to be determined is corrected and classified as “short” (step S109), and the process proceeds to step S110. As a result, the type of the cut length having a predetermined time (for example, the cut length between the third cut point and the fourth cut point) is corrected to “short”.

  On the other hand, when it is determined that there is an additional score (Yes in step S108), the process proceeds to step S110. In the case of proceeding to step S110, the third classification correction unit 110, for example, each classification of cut points and cut lengths after the event start time IN (including those that have been subjected to type correction in steps S105, S107, and S109). Is passed to the event end time extraction unit 111.

In step S110, the event end time extraction unit 111 refers to the appearance pattern (FIG. 4) in the appearance pattern storage unit 112, and sets the appearance pattern of the type classified in step S103 (including steps S105, S107, and S109). The corresponding event end time OUT is extracted.
For example, when the appearance pattern of the classified type to be extracted is the appearance pattern of FIG. 4A, the event end time extraction unit 111 sets the end of the cut length belonging to the “long” type as the event end Extracted as time OUT. Thereby, the end point of the cut length belonging to a certain type (for example, “long”) in a series of scenes composed of cut lengths having types of “short”, “medium”, and “long” is the event end time OUT. Extracted as

  Thereafter, the output unit 113 (extraction device 100) outputs the event end time OUT extracted by the event end time extraction unit 111 together with the event start time IN to an external device (for example, a computer). As a result, the event end time OUT of the sports video subject to digest viewing is output to the outside.

  As described above, according to the present embodiment, the extraction apparatus 100 includes the cut point detection unit 102 that detects a cut point after the event start time IN from the sports video in the sports video storage unit 101, and the cut point detection unit 102. Corresponding to the appearance pattern of the classified type by referring to the cut length classification unit 104 that classifies the type of cut length between the detected cut points according to the short length of the cut length and the appearance pattern storage unit 112 An event end time extraction unit 111 that extracts the event end time OUT to be output, and an output unit 113 that outputs the event end time extracted by the event end time extraction unit 111 and the event start time IN. Thereby, after making the appearance pattern of the appearance pattern storage unit 112 a series of video scenes, the end point (or start point) of a cut length having a certain time is extracted from the video scene as the event end time OUT. The Accordingly, it is possible to accurately extract the event end point of the sports video that is the subject of the digest viewing.

The present invention is not limited to the above-described embodiment, and the system configuration (including data configuration) and processing procedure may be changed without departing from the spirit of the present invention. For example, in FIG. 5, the case where each determination process in steps S104, S106, and S108 is performed has been described. However, in the classification process in step S103, the classified cut lengths are specified in advance (for example, 2.0). If there is no plus or minus α), the process may skip from step S103 to step S110.
In FIG. 5, the case where step processing is performed in the order of steps S <b> 104, S <b> 106, and S <b> 108 has been described. However, these orders may be reversed or changed to be changed.
Furthermore, in steps S104 to S109 in FIG. 5, the case where the first classification correction unit 108, the second classification correction unit 109, and the third classification correction unit 110 perform processing has been described. One classification correction unit may perform processing, or two arbitrary classification correction units may be combined to perform processing.

  Moreover, although the extraction apparatus 100 has been described with respect to a single computer, it may be configured to perform distributed processing using a plurality of computers. Furthermore, the extraction apparatus 100 may be realized by LSI (Large Scale Integration) or may be made into one chip.

  Furthermore, although the case where the sports video is a baseball broadcast video has been described, for example, if the sports video includes a telop that can determine the type of cut length, it can be applied to sports video such as American football and basketball. Also good.

It is a block diagram which shows the structural example of the event end point extraction apparatus in embodiment of this invention. It is a figure which shows an example of the sports image | video of the sports image | video memory | storage part of FIG. It is a figure which shows an example of the classification reference | standard of the classification reference | standard memory | storage part of FIG. It is a figure which shows an example of the appearance pattern of the appearance pattern memory | storage part of FIG. It is a figure which shows an example of the process sequence of the event end point extraction apparatus of FIG.

Explanation of symbols

DESCRIPTION OF SYMBOLS 100 Event end point extraction apparatus 101 Sports video memory | storage part 102 Cut point detection part 103 Cut point memory | storage part 104 Cut length classification | category part 105 Classification reference | standard storage part 106 Telop detection part 107 Telop storage part 108 1st classification correction | amendment part 109 2nd Classification correction unit 110 Third classification correction unit 111 Event end time extraction unit 112 Appearance pattern storage unit 113 Output unit

Claims (10)

A sports video storage unit that stores the sports video for which the event start point has been determined;
From the sports video of the sports video storage unit, a cut point detection unit that detects a cut point after the event start point,
A cut length classifying unit that classifies the type of cut length between the cut points detected by the cut point detecting unit according to the short length of the cut length;
An appearance pattern storage unit that stores an appearance pattern of the cut length type in association with an event end point;
An event end point extraction unit that refers to the appearance pattern storage unit and extracts an event end point corresponding to an appearance pattern of the classified type;
An output unit that outputs the event end point and the event start point extracted by the event end point extraction unit;
An event end point extraction device. As types of the cut length, there are a first type, a second type, and a third type in the order of short cut length,
As the appearance pattern of the appearance pattern storage unit, a pattern in which the third type first appears from the event start point is stored, and the end point of the cut length belonging to the third type corresponds as the event end point Attached,
The event end point extraction device according to claim 1. As types of the cut length, there are a first type, a second type, and a third type in the order of short cut length,
As the appearance pattern in the appearance pattern storage unit, a pattern in which the second type appears continuously from the event start point to the first appearance of the third type is stored. The end point of the cut length belonging to the second type is associated as the event end point.
The event end point extraction device according to claim 1. The sports video of the sports video storage unit includes a telop,
A telop detection unit for detecting the telop included in the sports video of the sports video storage unit;
A telop storage unit for storing the telop detected by the telop detection unit;
When the cut length classified by the cut length classification unit has a predetermined time, the classification type of the cut length is corrected to a preset type based on the display pattern of the telop in the telop storage unit A classification correction unit that classifies
The event end point extraction device according to claim 1. If the sports video is a baseball video,
As the display pattern of the telop, a display pattern of a specific telop that suggests a strikeout video scene is preset,
When the classification correction unit corrects the classification type of the cut length, if the specific telop display pattern suggesting the strikeout video scene is not in the cut length, the classification correction unit corrects the classification type of the cut length.
The event end point extraction device according to claim 4. If the sports video is a baseball video,
As the display pattern of the telop, a display pattern of a specific telop that suggests a video scene of changing batter is preset,
When the classification type correction unit corrects the classification type of the cut length, if the display pattern of the specific telop suggesting the video scene of the batter change is not in the cut length, the classification type correction unit corrects the classification type of the cut length. ,
The event end point extraction device according to claim 4. If the sports video is a baseball video,
As the display pattern of the telop, a display pattern of a specific telop that suggests an additional video scene is preset,
The classification type correction unit, when correcting the classification type of the cut length, corrects the classification type of the cut length if the display pattern of the specific telop that suggests the video scene to which the score is added is not in the cut length. ,
The event end point extraction device according to claim 4. An event end point extraction method used in a computer including a sports video storage unit that stores a sports video for which an event start point has been determined,
The computer
Detecting a cut point after the event start point from the sports video of the sports video storage unit;
Classifying the type of cut length between the detected cut points according to the short length of the cut length;
Extracting an event end point corresponding to an appearance pattern of the classified type with reference to an appearance pattern storage unit that stores an appearance pattern of the cut length type in association with an event end point;
Outputting the extracted event end point and the event start point;
Event end point extraction method. The sports video of the sports video storage unit includes a telop,
The computer
Detecting the telop included in the sports video in the sports video storage unit;
When the cut length classified in the cut length classification unit has a predesignated time, correcting the classification type of the cut length based on the detected display pattern of the telop and classifying the type; Further including
The event end point extraction method according to claim 8. An extraction program for causing a computer to execute the event extraction method according to claim 8 or 9.

Images