JP2005182402A - Field area detection method, system therefor and program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To automatically specify an imaging area (an area of an athletic field present inside an imaging screen) from a wide-angle image in a soccer video or the like. <P>SOLUTION: A field template generation means 110 generates a field template wherein a field line is arranged as a wire frame. A projection parameter group generation means 120 varies respective projection parameters independently of each other to generate a projection parameter group. A perspective conversion means 130 perspectively converts the field template from world coordinates into screen coordinates by use of each of the projection parameter group to obtain a plurality of template projection images. An image matching means 150 compares a binary field image with each of the plurality of template projection images to find a best-matched template projection image. A field detection means 160 finds a maximum coincident parameter, and detects the athletic field area inside a field image on the basis thereof. <P>COPYRIGHT: (C)2005,JPO&NCIPI

Description

  The present invention relates to a field area detection method, a system thereof, and a program. More specifically, the present invention relates to a field area detection method and system for automatically detecting a competition field area in video data obtained by photographing a sports competition. And the program.

  Many sports programs are broadcast on cable TV, digital TV broadcasting, terrestrial TV, etc., especially on cable TV and digital TV, etc., and the number of channels for these sports programs is enormous. Yes. With the widespread use of high-speed networks, a large amount of these sports videos can be viewed online. There is a need for a system that can efficiently search for scenes desired by users from such a vast video library. Especially, soccer is popular and video data is huge. Needs are high. Conventionally, a player, a ball, and a corner area are detected from a video of a soccer game (referred to as a “soccer video”, which is taken for the purpose of TV broadcasting), and the corner kick scene is detected based on their positional relationship. There have been proposed a method for realizing extraction, a method for detecting a player from a video of a camera that captures a fixed range on the field, and calculating a range of power of each team. However, each method has many mistakes to be solved, and is merely a proposal and has not been put into practical use.

  As a conventional technique, there is a method of extracting a corner kick scene from a soccer video (see Non-Patent Document 1). This is a technique in which players, balls, and corner areas are detected from a video of a soccer game (hereinafter referred to as soccer video), and a corner kick scene is extracted from their positional relationship. Since this conventional method detects the corner area based on the “goal post”, when the goal post is concealed by the player, the corner area (that is, the competition field area in the photographed image) cannot be detected, and the corner area is not detected. There is a problem that the kick scene cannot be detected.

As another conventional technique, a method of detecting the position of a player from a video of a camera that captures a fixed range on the field and specifying the power range of each team (see Non-Patent Document 1). In this conventional method, since the video of a camera that captures a fixed range on the field is targeted, video including panning and zooming (that is, a competition field region (ground, It is very difficult to apply to images including changes in the coat, pitch, etc.). In general, when a soccer video is captured for the purpose of TV broadcasting, it is inevitable to include video effects such as panning and zooming. Therefore, this method cannot be applied to TV relay video.
Maruo Jiro, Yoshio Iwai, Masahiko Taniuchi, Tomio Echigo, Shunichi Iizaku, “Extraction of specific video events from soccer video” (Science Technical Report I, IE99-17, PRMU99-41, MVE99-37, Ju1y 1999. ) Tsuyoshi Tsuji and Junichi Hasegawa “Characteristics and Application for Group Behavior Analysis in Team Sports” (The IEICE Transactions D-II, vol.J81-DII, pp. 1802-1181, August 1998.)

  In order to put the above-described conventional methods into practical use for the purpose of indexing the video acquired for television broadcasting and analyzing the content of the game, a wide-angle image is captured from a wide-angle image in a soccer video or the like. A method for automatically identifying the field area) is required. That is, in the case where the movement of the camera position and the shooting effect (such as panning and zooming) are included, the above-described conventional technique cannot be applied. In addition, in normal sports broadcast video, there are many scenes where it is difficult to “specify the competition field area”. In order to use the above-described conventional technology, for example, a human observes a photographed image and a wide-angle image is displayed. A method for identifying a competition field area by searching for an image and manually indicating on the screen several points (or lines) defining the competition field as control points (for example, four corner points of the competition field). However, it was difficult to automate because of the great human labor.

The field region detection method according to the present invention includes:
A field area detection method for detecting a competition field area,
A field template creation step for creating a field template in which the field line is arranged as a wire frame in world coordinates based on a standard that defines a field line of a competition field;
Projection parameter group generation step for generating a projection parameter group that includes projection parameters that can generate different projection images by independently changing parameters included in projection parameters used for perspective transformation from the world coordinates to the screen coordinates. When,
Using each of the projection parameter groups, the field template is perspective-transformed from the world coordinates to the screen coordinates to obtain a plurality of template projection images;
A binarization step of extracting a field line by converting a field image in which the competition field is photographed into a binary field image;
An image matching step of comparing the binary field image with each of the plurality of template projection images to obtain a template projection image that most closely matches;
A field detection step of setting a projection parameter of the most matched template projection image as a maximum matching parameter and detecting (calculating) a competition field region in the field image using a predetermined formula based on the maximum matching parameter;
It is a method including.
According to the present invention, the competition field area in the video data of the sport that competes in the competition field defined by several lines represented by soccer (or rugby, basketball, American football, etc.) is manually inserted. Automatic detection with high accuracy (ie, based on the field lines in the image, determine where the field area occupies in the shooting screen, and detect where this field area exists in the entire competition field) Be possible). Obtaining the imaging region automatically in this way is optimal as a pre-processing for other video analysis methods and can be widely applied. For example, it is possible to implement the present invention as a pre-processing of sports image database such as soccer or analysis of game content, and the processing of the present invention realizes indexing efficiently and with high accuracy, It becomes possible to analyze the game content smoothly.

The field region detection method according to the present invention includes:
Including the step of performing at least one expansion process on the binary field image,
It is characterized by that.
According to the present invention, noise such as dust and blur on the field line and various pixel noises mixed during binarization are reduced to reduce the field line (for example, goal line, half in soccer) Detection error of a line or a touch line) can be significantly reduced, and a region can be detected with higher accuracy.

Furthermore, the field region detection method according to the present invention includes:
Including performing a closing process (at least one expansion process and at least one contraction process) on the binary field image,
It is characterized by that.
According to the present invention, noise such as dust and blur on the field line and various pixel noises mixed at the time of binarization are reduced to further significantly reduce the detection error of the field line constituting the competition field, It becomes possible to detect the region with higher accuracy. If the field line on the screen is too thin, or the aspect ratio is horizontally long, such as a TV image, the horizontal line may disappear, but the line on the screen is sufficiently thick. This configuration works effectively.

Furthermore, the field region detection method according to the present invention includes:
A storage step of storing the field image in an image database in a format associated with camera shooting information;
A wide-angle image detection step of reading out the field image and camera shooting time information associated with the field image from the image database, and detecting the field image including a wide-angle image based on the camera shooting time information;
It is characterized by including.
Since the present invention is a technique for detecting an area based on a field line (goal line or side line) in an image, it is most accurate and effective when applied to an image over a wide range of the competition field. In recent digital video data, various data at the time of camera shooting such as focal length, image plane size, angle of view, subject distance, etc. may be held together, and images using this camera shooting data can be used. By automatically detecting an image photographed at a wide angle from the database, it becomes possible to detect a region more accurately and effectively. For example, since the angle of view can be obtained based on information on the focal length and the image plane size, only a wide-angle image can be extracted based on the obtained angle of view.

Furthermore, the field region detection method according to the present invention includes:
A storage step of storing the field image in an image database in a format associated with camera shooting information;
A step of reading out the field image and camera shooting time information associated therewith from the image database, and limiting the projection parameter group based on the camera shooting time information;
It is characterized by including.
According to the present invention, for example, each parameter group can be taken based on camera shooting information such as the relative position of the camera with respect to the competition field, shooting angle, field angle, focal length, image plane size, subject distance, and the like. Since it can be limited within the range, it is possible to remarkably reduce the perspective transformation process that consumes a large amount of computing power and the subsequent matching process, and the overall detection process can be speeded up.

As described above, the solution of the present invention has been described as a method. However, the present invention can also be realized as a system (apparatus), a program, and a storage medium that records the program. It should be understood that these are included in the scope of the invention.
For example, when the present invention is realized as a system, the field region detection system according to the present invention is:
A field area detection system for detecting a competition field area,
Field template creation means for creating a field template in which the field line is arranged as a wire frame in world coordinates based on a standard that defines the field line of the competition field;
Projection parameter group generation means for generating a projection parameter group including projection parameters that can independently change parameters included in projection parameters used when perspective-transforming from the world coordinates to the screen coordinates, and that can generate different projection images. When,
Using each of the projection parameter groups, perspective transformation means for perspective transformation of the field template from the world coordinates to the screen coordinates to obtain a plurality of template projection images;
Binarizing means for converting a field image in which the competition field is photographed into a binary field image and extracting a field line;
Image matching means for comparing the binary field image and each of the plurality of template projection images to obtain a template projection image that most closely matches;
Field detection means for setting a projection parameter of the most matched template projection image as a maximum matching parameter and detecting (calculating) a competition field region in the field image based on the maximum matching parameter;
Is a field region detection system including:

Moreover, the field area detection system according to the present invention includes:
Means for subjecting the binary field image to expansion processing;
Means for applying a closing process to the binary field image;
Storage means for storing the field image in an image database in a format associated with camera shooting information;
Wide-angle image detection means for reading out the field image and camera shooting time information associated with the field image from the image database, and detecting the field image including a wide-angle image based on the camera shooting time information;
It is characterized by including.
Furthermore, the field area detection system according to the present invention comprises:
Storage means for storing the field image in an image database in a format associated with camera shooting information;
Limiting means for reading out the field image and camera shooting time information associated therewith from the image database, and limiting the projection parameter group based on the camera shooting time information;
It is characterized by including.

The program according to the present invention is
A program for causing a computer to execute a field area detection method for detecting a competition field area,
A field template creation step for creating a field template in which the field line is arranged as a wire frame in world coordinates based on a standard that defines a field line of a competition field;
Projection parameter group generation step for generating a projection parameter group that includes projection parameters that can generate different projection images by independently changing parameters included in projection parameters used for perspective transformation from the world coordinates to the screen coordinates. When,
Using each of the projection parameter groups, the field template is perspective-transformed from the world coordinates to the screen coordinates to obtain a plurality of template projection images;
A binarization step of extracting a field line by converting a field image in which the competition field is photographed into a binary field image;
An image matching step of comparing the binary field image with each of the plurality of template projection images to obtain a template projection image that most closely matches;
A field detection step of setting a projection parameter of the most matched template projection image as a maximum matching parameter, and detecting (calculating) a competition field region in the field image based on the maximum matching parameter;
Including programs.

Furthermore, the program according to the present invention is:
Performing a dilation process on the binary field image;
Applying a closing process to the binary field image;
A storage step of storing the field image in an image database in a format associated with camera shooting information;
A wide-angle image detection step of reading out the field image and camera shooting time information associated with the field image from the image database, and detecting the field image including a wide-angle image based on the camera shooting time information;
It is characterized by including.
Furthermore, the program according to the present invention is:
A storage step of storing the field image in an image database in a format associated with camera shooting information;
A step of reading out the field image and camera shooting time information associated therewith from the image database, and limiting the projection parameter group based on the camera shooting time information;
It is characterized by including.

Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.
FIG. 1 is a block diagram showing a basic configuration of a field area detection system according to the present invention. As shown in the figure, the field region detection system 100 of the present invention includes a field template creation unit 110, a projection parameter group generation unit 120, a perspective transformation unit 130, a binarization unit 140, an image matching unit 150, and a field detection unit 160. With
The field template creation means 110 creates a field template in which the field lines are arranged as wire frames in world coordinates based on a standard that defines the field lines of the competition field. The projection parameter group generation means 120 changes the parameters included in the projection parameters used when perspective-transforming from the world coordinates to the screen coordinates, and includes projection parameters that can generate different projection images. Is generated. The perspective conversion unit 130 performs perspective conversion of the field template from the world coordinates to the screen coordinates using each of the projection parameter groups, and obtains a plurality of template projection images. The binarization means 140 reads a field image in which the competition field is photographed from an image database, converts the read image into a binary field image, and extracts a field line. The image matching means 150 compares the binary field image and each of the plurality of template projection images to obtain a template projection image that most closely matches. The field detection unit 160 uses the projection parameter of the most matched template projection image as a maximum matching parameter, and detects a competition field region in the field image based on the maximum matching parameter. The detected field area is associated with the image, passed to another image detection system (not shown), and used for subsequent processing.

FIG. 2 is a flowchart illustrating a field region detection method according to the present invention.
The template creation step S110 creates a field template in which the field lines are arranged as wire frames in world coordinates based on a standard that defines the field lines of the competition field. In the projection parameter group generation step S120, parameters included in the projection parameters used when performing perspective transformation from the world coordinates to the screen coordinates are independently changed, and projection parameter groups each including a projection parameter that can generate different projection images. Is generated. The perspective transformation step S130 perspectively transforms the field template from the world coordinates to the screen coordinates using each of the projection parameter groups to obtain a plurality of template projection images. In the binarization step S140, in order to extract a field line, a field image in which the competition field is photographed is read from an image database, and the read image is converted into a binary field image. The image matching step S150 compares the binary field image with each of the plurality of template projection images to obtain a template projection image that most closely matches. In the field detection step S160, the projection parameter of the best matching template projection image is set as the maximum matching parameter, and the competition field region in the field image is detected based on the maximum matching parameter.

Soccer is particularly popular in sports broadcast programs, and the important scenes in soccer video are shoot scenes where scores can be scored, but such scenes are very few overall, such scenes It is preferable to apply the detection technique of the present invention to pre-processing of various techniques for extracting a large amount of video data. Although the present invention can be applied to various sports, in light of the above points, a soccer video is taken as an example of a sports competition and will be described as a method / system for detecting a soccer competition field area of a soccer video.
First, when analyzing the content of a game in a soccer video, it is necessary to specify an imaging region in advance from an image (hereinafter referred to as a wide-angle image) in which a field that occupies a large amount in the soccer video is captured. The present invention realizes field region detection in a video shot without limiting the camera position and shooting effect (such as pan and zoom).

The present invention specifies a shooting area in a soccer field from a wide-angle video in a soccer video. It has the following configuration for implementation.
1) Extract wide-angle video from soccer video.
2) Extract field lines from wide-angle images.
3) Using a field line as a wire frame model (hereinafter referred to as a field template), an image obtained by projecting the field template onto a plane by perspective transformation is obtained.
4) Compare the field line obtained in 2) with the image obtained in 3), and detect the projection parameter whose position is closest.
5) Use the obtained projection parameters to identify the field area captured.

According to the present invention, it is possible to identify a development field region in a wide-angle image that is indispensable for analyzing and indexing a soccer video. The present invention can be realized even when it includes a shooting effect (panning, zooming, etc.) generally performed in editing of a television image, and is versatile.
The present invention provides a technique for detecting a range of a field that has been captured from an image (hereinafter, a wide-angle image) in which the field is captured over a wide range by a camera that performs panning and zooming. In the present invention, a field line is used as a wire frame model (hereinafter referred to as a field template). An image obtained by projecting the field template onto a plane by perspective transformation and a field line extracted from the wide-angle image are compared, and a projection parameter with the closest position is detected. The region of the projected field can be specified using the obtained projection parameter.
In the method according to the present invention, an image obtained by projecting a field template onto a plane by perspective transformation is compared with an image obtained by extracting an image forming a field line from a wide-angle image for each pixel. Through the comparison, a projection parameter with the highest degree of matching of images is detected, and a photographed region is calculated using the value. Since the field template used in the present invention is composed of a plurality of lines, it is possible to detect an imaged region without being affected by concealment by the player. Each process will be described below.

Projection of field templates The Japan Football Association and the International Football Federation have adopted the fields shown in Figure 3 for national-scale tournaments (such as Wooled Cup, Olympics, World Youth Games, etc.) (Please refer to "Soccer Competition Rules, 1999/2000" (Japan Football Association Referee Committee, "Japan Football Association, pp, 6-10, 1999)"). Therefore, in the present invention, the model shown in FIG. 3 is adopted as the field template.

である。以降、上の座標変換によって得られた画像をテンプレート投影画像と呼ぶ。ただし、テンプレート投影画像において、ラインを1，それ以外を0で表現する。 In order to express the projected image of the field template, world coordinates x, y, z and screen coordinates X, Y are set as shown in A and B of FIG. Place the field template on the xy plane of world coordinates so that the touch line and goal line are parallel to the x and y axes, respectively, and the viewpoint is P _v (x _v , y _v , z _v ), and the projection direction is P _f (x _f , y _f , 0), and the distance from the viewpoint to the screen is represented by r. However, the straight line passing through P _v P _f is perpendicular to the screen. Also, the angle formed by the line segment l and the x axis projected from the line segment P _v P _f on the xy plane is θ, the angle formed by the line segment l and the line segment P _v P _f is the center point of the screen in world coordinates. _{P s (x s, y s} , z s), the length of the line segment P _v P _s is denoted by r '. At this time, perspective transformation from world coordinates to screen coordinates is expressed by equations (1) and (2).

It is. Hereinafter, the image obtained by the above coordinate conversion is referred to as a template projection image. However, in the template projection image, the line is represented by 1 and the others are represented by 0.

の範囲に限定できる。即ち、この式は、サッカースタジアムの外にカメラは設置されないこと、及び、サッカーフィールドの中にカメラは設置されないことを意味する。このように設定された範囲は、可能な設定値を網羅するよう十分に大きく設定されている。従って、競技場によるカメラの設置位置の違いにより、フィールド領域の検出結果に劣化が生じる危険性は少ない。投影方向はカメラの撮影角度を表しているので、カメラが常にフィールド内を撮影しているとすれば、フィールドの大きさの規格より

の範囲に限定される。スクリーンまでの距離ｒはカメラの画角に関係しているので、フィールドがスクリーン内に納まるように範囲を定める。このように投影パラメータの取り得る範囲を限定して、投影パラメータ群を構成する数を減少させれば、透視変換やマッチング処理などの計算負荷を顕著に低減でき、処理の迅速化、ひいては検出精度の向上も図れる。 The projection parameter P _v P _fr used for generating the template projection image can limit the range that each projection parameter can take based on the state of the camera estimated from the actual soccer video as follows. Viewpoint P _v is based on the fact that a camera that captures wide-angle images in soccer matches is installed near the center of the audience seat on the touchline side.

Can be limited to the range. That is, this equation means that no camera is installed outside the soccer stadium and no camera is installed in the soccer field. The range set in this way is set sufficiently large to cover the possible set values. Therefore, there is little risk that the detection result of the field area is deteriorated due to the difference in the camera installation position depending on the stadium. Since the projection direction represents the shooting angle of the camera, if the camera is always shooting in the field, the field size standard

It is limited to the range. Since the distance r to the screen is related to the angle of view of the camera, the range is determined so that the field fits within the screen. By limiting the range of projection parameters that can be taken in this way and reducing the number of projection parameter groups, the computational load such as perspective transformation and matching processing can be significantly reduced. Can be improved.

を満たす画素をフィールドラインとする。ただし、Th_Yは閾値であり、2／5π、πはマンセル表色系でそれぞれ5Y（yellow）,5BG（bluegreen）を表している。この処理により得られたラインを1、それ以外を0とした画像を以降、2値フィールド画像と呼ぶ。また、2値フィールド画像に膨張処理を施した画像（以下、膨張2値フィールド画像）を準備し、以降で行われるフィールド投影画像の比較の対象とすることにより、ラインの検出誤差の影響を軽減する。 Extraction of Field Line from Wide Angle Image Next, a method for extracting pixels constituting the field line from the wide angle image will be described.
In many competitions, the field line is white, and white is characterized by high brightness. Further, since the field line is often drawn on the green grass, the pixels constituting the field line on the image are a mixed color of white and green. Therefore, y (luminance) of the YIQ color system and the HSV color system (hue) are used.

Pixels that satisfy are defined as field lines. However, Th _Y is a threshold value, and 2 / 5π and π represent 5Y (yellow) and 5BG (bluegreen) in the Munsell color system, respectively. An image in which the line obtained by this processing is 1 and the others are 0 is hereinafter referred to as a binary field image. In addition, by preparing an image obtained by performing dilation processing on a binary field image (hereinafter, dilated binary field image) and making it a target for comparison of field projection images to be performed later, the influence of line detection errors is reduced. To do.

In limiting the scope as detected above the shooting range by comparison with the template projection image and the expansion binary field image, change Yv, Zv, x _f, δyz increments y _f, r the independently, δyv, δyf, in δr The template projection image and the expanded binary field image are compared for all combinations of values of the parameters. At this time, as an index for comparison, the number of pixels having a value of 1 in the image obtained by taking the logical product of the template projection image and the expanded binary field image is used (hereinafter referred to as template matching degree). In this specification _, P _v, P _f, r when the highest template matching degree is obtained by this processing will be hereinafter referred to as a maximum matching parameter. However, since the only projection parameter that changes due to panning and zooming of the camera is P _f, r, when detecting the maximum matching parameter for moving images, if P _v is calculated in the first frame, then P _{f ,} R only search. Assuming that the camera is always following the ball, the speed of the ball is 40 m / s at the time of shooting. Therefore, if P _f at an adjacent frame is known in the video sampled at Nfps, the 40 / Only the neighborhood of Nm needs to be searched. Similarly, the search range can be reduced by using the fact that r also changes little in adjacent frames.

式（10）および（11）にスクリーンの4隅の座標を代入し、xy平面上に撮影された領域を算出する。 Next, a method for calculating the area of the photographed field from the maximum matching parameter will be described. The rectangles with the four corners of the screen as vertices correspond to the shooting range. Therefore, the coordinates of the points on the xy plane projected onto these four points are obtained as follows. Transform equations (1) and (2) with Z = 0 to obtain an equation that converts from screen coordinates to world coordinates.

Substituting the coordinates of the four corners of the screen into equations (10) and (11), the area photographed on the xy plane is calculated.

In order to confirm the effectiveness of the present invention, the following experiment is performed and the results are shown. The experiment used a wide-angle image (360 x 240) in RGB format for about 14 seconds (137 frames) sampled at 10 fps. The threshold used in the experiment is Thy = 0.47, the number of expansion processes is 3, the search range of r is r∈ [800, 1300], P _v, P _{f, and} the step size of r are δy _z , δy _v , δy _f , Δr = 10. A part of the experimental results is shown in FIG. 5 (A is a wide-angle image, B is a projected image, and C is a diagram showing an imaging region detected by the present invention). From the figure, it can be confirmed that the field lines of the wide-angle image (A) and the template projection image (B) obtained by photographing the actual competition field match, and the photographed region (C) is correctly detected. When the correctness / incorrectness of the detection result of the verification system according to the present invention is visually determined, 131 frames out of 137 frames have been correctly detected. Therefore, according to the present invention, it has been found that the field region can be detected with considerably high accuracy without any manual work on the video data.

  The present invention is expected to be used in the broadcasting industry and sports video database industries that require sports video analysis and the like.

In the present specification, the principle of the present invention has been described in various embodiments. However, the present invention is not limited to the above-described embodiments, and many variations and modifications can be made by those skilled in the art. It should be understood that these variations and modifications are also included in the present invention.
For example, in the embodiment, the matching process is of the type compared for each pixel, but various matching processes as described below can be used in the present invention.
Modification 1 of matching process
For each line segment of the result of applying perspective transformation to the field template (the wire frame model is not converted into a bitmap image), a pixel with a value of 1 near the d pixel of the line segment in the binary field image The number is calculated, and the total sum of all line segments is set as the template matching degree.
Modification 2 of matching process
The existing line extraction method is applied to the binary field image, and the obtained line segment is compared with the projected image of the field template.

  In the above-described embodiment, the search range of the parameters is limited by assuming a general camera arrangement with the size of a general soccer competition stadium, but the camera shooting information automatically acquired by the camera at the time of shooting, If it is possible to obtain the camera shooting information added in step (1), it is possible to further limit the parameter search range, and to quickly detect the area. As the information at the time of camera shooting, for example, there are a relative position of the camera with respect to the competition field, shooting angle, angle of view, focal length, image plane size, subject distance, etc. It is possible to considerably limit the parameter search range.

It is a block diagram which shows the fundamental structure of the field area | region detection system by this invention. 5 is a flowchart illustrating a field region detection method according to the present invention. It is a top view which shows the field template corresponding to a soccer field. A is a diagram illustrating world coordinates x, y, and z that represent the projected image of the field template in three dimensions, and B is a diagram illustrating screen coordinates x and y that represent the projected image of the field template in two dimensions. A is a diagram showing a wide-angle image obtained by photographing an actual field, B is a diagram showing a template projection image, and C is a diagram showing an imaging region detected by the present invention.

Explanation of symbols

DESCRIPTION OF SYMBOLS 100 Field area | region detection system 110 Field template preparation means 120 Projection parameter group production | generation means 130 Perspective transformation means 140 Binarization means 150 Image matching means 160 Field detection means

Claims (12)

A field area detection method for detecting a competition field area,
A field template creation step for creating a field template in which the field line is arranged as a wire frame in world coordinates based on a standard that defines a field line of a competition field;
Projection parameter group generation step for generating a projection parameter group that includes projection parameters that can generate different projection images by independently changing parameters included in projection parameters used for perspective transformation from the world coordinates to the screen coordinates. When,
Using each of the projection parameter groups, the field template is perspective-transformed from the world coordinates to the screen coordinates to obtain a plurality of template projection images;
A binarization step of extracting a field line by converting a field image in which the competition field is photographed into a binary field image;
An image matching step of comparing the binary field image with each of the plurality of template projection images to obtain a template projection image that most closely matches;
A field detection step of setting a projection parameter of the most matched template projection image as a maximum matching parameter and detecting a competition field region in the field image based on the maximum matching parameter;
Field area detection method including: The field region detection method according to claim 1,
Including the step of dilating the binary field image,
A field region detection method characterized by the above. In the field area detection method according to claim 1 or 2,
A storage step of storing the field image in an image database in a format associated with camera shooting information;
A wide-angle image detection step of reading out the field image and camera shooting time information associated with the field image from the image database, and detecting the field image including a wide-angle image based on the camera shooting time information;
A field region detection method comprising: In the field area detection method according to claim 1 or 2,
A storage step of storing the field image in an image database in a format associated with camera shooting information;
A step of reading out the field image and camera shooting time information associated therewith from the image database, and limiting the projection parameter group based on the camera shooting time information;
A field region detection method comprising: A field area detection system for detecting a competition field area,
Field template creation means for creating a field template in which the field line is arranged as a wire frame in world coordinates based on a standard that defines the field line of the competition field;
Projection parameter group generation means for generating a projection parameter group including projection parameters that can independently change parameters included in projection parameters used when perspective-transforming from the world coordinates to the screen coordinates, and that can generate different projection images. When,
Using each of the projection parameter groups, perspective transformation means for perspective transformation of the field template from the world coordinates to the screen coordinates to obtain a plurality of template projection images;
Binarizing means for converting a field image in which the competition field is photographed into a binary field image and extracting a field line;
Image matching means for comparing the binary field image and each of the plurality of template projection images to obtain a template projection image that most closely matches;
Field detection means for setting a projection parameter of the most matched template projection image as a maximum matching parameter, and detecting a competition field region in the field image based on the maximum matching parameter;
Including field area detection system. The field region detection system according to claim 5,
Means for subjecting the binary field image to expansion processing;
A field region detection system characterized by that. The field region detection system according to claim 5 or 6,
Storage means for storing in the image database in a format in which camera field shooting time information is associated with the field image;
Wide-angle image detection means for reading out the field image and camera shooting time information associated with the field image from the image database, and detecting the field image including a wide-angle image based on the camera shooting time information;
A field region detection system comprising: The field region detection system according to claim 5 or 6,
Storage means for storing in the image database in a format in which camera field shooting time information is associated with the field image;
Limiting means for reading out the field image and camera shooting time information associated therewith from the image database, and limiting the projection parameter group based on the camera shooting time information;
A field region detection system comprising: A program for causing a computer to execute a field area detection method for detecting a competition field area,
A field template creation step for creating a field template in which the field line is arranged as a wire frame in world coordinates based on a standard that defines a field line of a competition field;
Projection parameter group generation step for generating a projection parameter group that includes projection parameters that can generate different projection images by independently changing parameters included in projection parameters used for perspective transformation from the world coordinates to the screen coordinates. When,
Using each of the projection parameter groups, the field template is perspective-transformed from the world coordinates to the screen coordinates to obtain a plurality of template projection images;
A binarization step of extracting a field line by converting a field image in which the competition field is photographed into a binary field image;
An image matching step of comparing the binary field image with each of the plurality of template projection images to obtain a template projection image that most closely matches;
A field detection step of setting a projection parameter of the most matched template projection image as a maximum matching parameter and detecting a competition field region in the field image based on the maximum matching parameter;
Including programs. The program according to claim 9,
Including the step of dilating the binary field image,
A program characterized by that. The program according to claim 9 or 10,
A storage step of storing the field image in an image database in a format associated with camera shooting information;
A wide-angle image detection step of reading out the field image and camera shooting time information associated with the field image from the image database, and detecting the field image including a wide-angle image based on the camera shooting time information;
The program characterized by including. The program according to any one of claims 9 and 10,
A storage step of storing the field image in an image database in a format associated with camera shooting information;
A step of reading out the field image and camera shooting time information associated therewith from the image database, and limiting the projection parameter group based on the camera shooting time information;
The program characterized by including.

Images