JPH1015152A - Recognition device for pieces on japanese chess board - Google Patents

Abstract

PROBLEM TO BE SOLVED: To easily recognize the existence or non-existence of Japanese chess pieces and raise a judgement treatment rate for the existence or non-existence of the pieces by preparing a histogram on the basis of an inter-image computation result between a label image and an edge extracted image, and applying a threshold value to the histogram. SOLUTION: An edge extracted image is generated from the images of pieces on a Japanese chess board picked up via a CCD camera 1, using the first luminance value of square lines on the board and characters on each piece, and the second luminance value of the bare surface of the board. The preset inter-image calculation is then made between a label image at a position corresponding to the inside of each square on the board in the edge extracted image, and the edge extracted image for obtaining the image where only a picture element having the first luminance value among picture elements having the first and the second luminance values becomes a luminance value for a corresponding label in zones corresponding to each label on the edge extracted image. At the same time, a histogram is prepared to show the frequency of picture elements for the luminance value, based on the image so obtained, and a frequency for each luminance value is compared with a threshold value to judge the existence or non-existence of pieces.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an invalid area extracting apparatus for extracting an area where obstacles such as hands are present on a shogi board based on the captured image of the pieces on the shogi board, and to an invalid area extracting apparatus for extracting the area on the shogi board. And an apparatus for recognizing the presence or absence of a piece on a shogi board.

[0002]

2. Description of the Related Art The applicant of the present invention employs an invalid area extracting means for extracting an invalid area in which a shogi board is hidden by an obstacle based on a captured image of a piece on the shogi board. A piece presence / absence recognition means for recognizing the presence or absence of a piece for each cell, a change area extraction means for extracting an area in which the piece presence / absence state has been changed based on the recognition result by the piece presence / absence recognition means, and an area where the piece presence / absence state has been changed are extracted. Developed a game record recording device equipped with a recording means for recording the type and position of a changed piece when it was done and recording the details of the change of the piece in a storage device for each move based on the type and position of the specified piece. And filed a patent application. However, at the time of filing the present application, it has not been published yet.

The piece presence / absence recognizing means in the game record recording apparatus developed by the present applicant uses the
A total sum of white pixels (hereinafter, referred to as a histogram value for each square) is calculated for each square. Then, by comparing the histogram value for each cell with the threshold value, it is determined whether or not a piece exists in each cell.

[0004]

SUMMARY OF THE INVENTION An object of the present invention is to provide an apparatus for recognizing the presence / absence of a piece on a shogi board which can improve the processing speed.

[0005]

SUMMARY OF THE INVENTION A piece presence / absence recognition device on a shogi board according to the present invention recognizes the presence / absence of a piece for each square based on a captured image of the pieces on the shogi board. In, from the captured image of the pieces on the shogi board,
The luminance value of the line separating the squares on the shogi board and the character drawn on each piece is the first luminance value, and the luminance value of the background inside the squares on the shogi board is different from the first luminance value. Edge extraction image generating means for generating an edge extraction image which has a label at a position corresponding to the inside of each square on the shogi board on the edge extraction image, and a luminance value of a portion other than the label is a third luminance value A predetermined image between the label image whose luminance value in each label is different from the second luminance value and the third luminance value and which is different for each label, and the edge extracted image. By performing the operation, in the region corresponding to each label on the edge extraction image, only the pixel of the first luminance value corresponds to the luminance value of the pixel of the first luminance value and the pixel of the second luminance value. Inter-image rendering to obtain an image that is the luminance value of the label Means, means for creating a histogram representing the frequency of pixels with respect to the luminance value based on the image obtained by the inter-image calculation means, and comparing the frequency for each of the obtained luminance values with a threshold value, wherein the frequency is determined. Threshold value processing means for determining that there is a piece in a cell in which a label corresponding to a luminance value equal to or greater than the value is present, and determining that there is no piece in a cell in which a label corresponding to a luminance value whose frequency is smaller than the threshold value is present; It is characterized by having.

[0006] The first luminance value, the second luminance value, the third luminance value, and the luminance value of each label are represented by, for example, a binary value of a predetermined number of bits. For example, the first luminance value is
All bits are set to a binary value represented by one,
The luminance value and the third luminance value are set to a binary value in which all bits are represented by 0, and the luminance value of each label is 2 in which all bits are larger than a binary number of 0 and all bits are 1 Binary values that are different from each other are set within a range of a decimal number or less. When each luminance value is set in this way, the inter-image calculation means sets the A between the label image and the edge-extracted image.
An ND operation is performed.

According to the apparatus for recognizing pieces on a shogi board according to the present invention, a histogram is created based on an inter-image operation result between a label image and an edge extracted image, and threshold processing is performed on the histogram. By simply performing the operation, the presence or absence of a frame is recognized, so that the processing speed is improved.

[0008]

Embodiments of the present invention will be described below with reference to the drawings.

FIG. 1 shows the appearance of a game record recording apparatus.

The game record recording device is a CCD for imaging a shogi board.
The camera includes a camera 1 and a personal computer (hereinafter, referred to as a host) 19. The host 19 has an internal monitor 22 for displaying a game record image. Further, the host 19 is equipped with an image processing board 30 for performing image processing on an image captured by the CCD camera 1.
The external monitor 5 is connected to the image processing board 30.

FIG. 2 shows an electric configuration of the game record recording apparatus, and in particular, an image processing board 3 mounted on the host 19.
0 shows a detailed configuration.

1 is a CCD for taking in an image of a shogi board
Camera. An analog image signal captured by the CCD camera 1 is converted by the A / D converter 2 into a digital image signal.

The image bus 6 is connected to a frame memory 7, an image processing unit 11, and an image memory 13.

A digital image obtained by the A / D converter 2 is stored in the frame memory 7. The first memory control unit 8 and the video control unit 9 control writing of an image to the frame memory 7 in synchronization with the output of the CCD camera 1. The image stored in the frame memory 7 (hereinafter, referred to as a captured image) is displayed on the external monitor 5 via the D / A converter 4.

The image processing section 11 performs image processing such as filter processing, binarization processing, histogram processing, and template matching.

The image memory 13 stores the results of various processes by the image processing section 11 and the template of the frame used for template matching. The image memory 13
An area (hereinafter, referred to as a queue) for storing and holding an image captured by the CCD camera 1 by the interrupt processing is provided in the area. In this example, the queue has an area in which ten captured images can be stored. The image memory 13 is controlled by the second memory control unit 14.

The CPU 16 is a central processing unit mounted on the image processing board 30. In the CPU bus 15,
In addition to the CPU 16, the first and second memory control units 8 and 14, the frame memory 7, the image memory 13, and the image processing unit 11, a processor EPROM 18 and a host interface 17 are connected.

The EPROM 18 stores a program for the CPU 16, that is, a control program for the image processing board 30. The host interface 17 exchanges data between the CPU bus of the host 19 and the CPU bus 15 of the image processing board 30.

The host 19 has a storage device 23 for storing the program, the game record, and other necessary data. The host 19 issues commands to the image processing board 30 and performs various processes related to a user interface and the like.

FIG. 3 shows the procedure of the main processing of the game record recording apparatus.

First, before a game is started, an initialization process is performed (step 1). In the initialization processing, area setting of the shogi board, creation of a template for each piece, and the like are performed.

At the start of the game, information on the type and position of each frame corresponding to the arrangement of the frame at the start of the game is stored in the storage device 23 of the host 19, and the internal monitor 22 of the host 19 stores the information at the time of the start of the game. The arrangement image of the pieces is displayed.

When the game starts (step 2), an image input process is executed (step 3). In this image input processing, when no captured image is stored in the queue, a game image is captured from the CCD camera 1. When captured images are stored in the queue, the game image is captured from the queue. The details of this image input processing will be described later.

Next, the frame movement detection and moving frame identification processing of steps 4 to 9 are executed. In the frame movement detection and moving frame identification processing, first, an invalid area extraction processing is performed based on the captured image (step 4). The invalid area is an area where an obstacle such as a player's hand exists in the captured image. Details of the invalid area extraction processing will be described later.

Thereafter, a frame presence / absence recognition process is performed (step 5). That is, the presence or absence of a piece is determined for each square in an area (effective area) excluding the invalid area on the chess board.
The details of the frame presence / absence recognition processing will be described later.

Next, a process of extracting a change area of the frame presence state is performed (step 6). That is, in the area (effective area) excluding the invalid area on the shogi board, an area where the state of the presence or absence of the piece has changed from the previous game record is extracted.

If an area in which the state of the presence or absence of a frame has been changed is extracted (YES in step 7), a process of specifying a moving frame is performed (step 8). In other words, the type of the moved piece and the position of the moved piece are specified by the shogi rule and the pattern matching.

If the type and position of the moved piece could not be specified in the moving piece specifying process in step 8 (including the case where the movement of the piece was determined to be invalid (rule violation)) (NO in step 9) ), Returning to step 3 to perform image input processing.

If the type and position of the moved piece can be specified in the moving piece specifying process in step 8 (YES in step 9), the game record is recorded in the storage device 23 in the host 19 (step 10). .

If there is no input for ending the record of the game record (NO in step 11), the process returns to step 3 to perform the image input processing. If there is an input to end the game record recording (YES in step 11), the game record recording process ends.

If it is determined in step 6 that an area in which the presence or absence of a frame has changed is not extracted (step 7).
NO), it is determined that the game record has not changed, and the routine proceeds to step 12.

In step 12, it is determined whether or not to check whether the latest game record stored in the storage device 23 of the host 19 is correct. In this example, step 1
The number of transitions to 2 is counted, and when the counted number reaches a predetermined value, it is determined that a check is to be performed, and the count value is set to 0.

When it is determined that no check is performed, the process proceeds to step 11. If it is determined that the check is to be performed, the latest game record stored in the storage device 23 is stored based on the latest game record stored in the storage device 23 of the host 19 and the currently captured image. It is checked whether it is correct (step 13). In this check processing, a timer for starting the interrupt processing is started. For more information about the check process,
It will be described later.

As a result of the check processing, when it is determined that the latest game record stored in the storage device 23 of the host 19 is correct (YES in step 14), the process proceeds to step 11.

As a result of the check processing, when it is determined that the latest game record stored in the storage device 23 of the host 19 is incorrect (NO in step 14), a recovery process for recognizing a correct game record is performed. Executed (step 1
5). The details of the recovery process will be described later.

When the correct game record is recognized by the recovery process, the process proceeds to step 10 and the recognized correct game record is stored in the storage device 23 of the host 19.

FIG. 4 shows a detailed procedure of the invalid area extraction processing in step 4 of FIG.

FIG. 5 shows an example of a captured image.
The player's hand (obstacle) appears on the shogi board of the captured image. The captured image has 0 to 255 gradations (black: 0,
White: 255). As shown in FIG. 5, it is assumed that the lines dividing the squares on the chess board are black, and the characters drawn on each piece are also black. In the captured image, the luminance value of the player's hand is lower (darker) than the luminance value of the shogi board against the background.

Hereinafter, the invalid area extraction processing procedure will be described with reference to the captured image of FIG. 5 as an example. First, a noise removal process is performed on a captured image using a low-pass filter (step 101). An edge extraction process is performed on the image from which noise has been removed by the noise removal process using a high-pass filter (step 10).
2). Then, a binarization process is performed on the obtained image (step 103). As a result, as shown in FIG. 6, the edges, that is, the lines dividing the squares on the shogi board, the characters drawn on each piece and the outline of the hand are white (8-bit binary value representing 255 in decimal). "1111111
1 "), and the portion other than the edge portion, that is, the inside of the square on the shogi board and the inside of the outline of the hand are black (0 in decimal number).
Is obtained (hereinafter, referred to as an edge-extracted image) which is an 8-bit binary value “00000000”).

Next, an AND operation is performed between the first label image and the edge extracted image (step 10).
4). FIG. 7 shows a first label image. The first label image is an image having rectangular labels A1 to A100 at the respective intersections of the horizontal and vertical lines that form the lines dividing the cells of the chess board.

In the first label image, the labels A1 to A1
Portions other than A100 are black (8-bit binary value “00000000” representing decimal 0). Each label A
The luminance values in 1 to A100 are different 8-bit binary values representing decimal numbers 1 to 100. That is, the luminance value of the label A1 is an 8-bit binary value representing a decimal number 1, the luminance value of the label A2 is an 8-bit binary value representing a decimal number 2, and the luminance value of the label A100 is It is an 8-bit binary value representing 100 in decimal.

When an AND operation is performed between such a first label image and an edge-extracted image, the luminance value becomes 0 in areas other than the labels A1 to A100 in the edge-extracted image. In addition, as shown in FIG. 8A, among the areas corresponding to the labels A1 to A100 in the edge extraction image, as shown in FIG. Is the brightness value of the corresponding label Ai, and the brightness value of the portion other than the line dividing the cells in the area is 0. Therefore, A
In the image after the ND operation, among the areas corresponding to the labels A1 to A100 in the edge extracted image, in the area where no obstacle such as the hand is present, there are relatively many pixels of the luminance value of the corresponding label Ai. Will do.

On the other hand, each label A in the edge extracted image
As shown in FIG. 8B, among the areas corresponding to 1 to A100, in the area in which all the lines dividing the squares by obstacles F such as hands are covered, the brightness values in the area are all 0. Becomes Therefore, in the image after the AND operation, in the area corresponding to each of the labels A1 to A100 in the edge extraction image, in the area where all the lines separating the squares by the obstacle F such as the hand are covered by the corresponding label A
The pixel having the luminance value of i does not exist at all.

Each label A in the edge extracted image
As shown in FIG. 8C, among the areas corresponding to 1 to A100, in the area where a part of the line dividing the cell in the area is covered by an obstacle F such as a hand, Among the lines dividing the squares, the luminance value of the portion not covered by the obstacle F is the luminance value of the corresponding label Ai, and the luminance value of the other portions is 0. Therefore,
In the image after the AND operation, of the regions corresponding to the labels A1 to A100 in the edge extracted image, in the region where a part of the line dividing the cell in the region is covered by an obstacle such as a hand, the corresponding region is used. There will be a few pixels with the luminance value of the label Ai.

When the AND operation is performed between the first label image and the edge-extracted image, the frequency of the pixel with respect to the luminance value is represented based on the obtained image as shown in FIG. A histogram is created (step 1
05).

Then, first threshold processing is performed based on the created histogram (step 10).
6). That is, a label position corresponding to a luminance value whose histogram value (frequency) is smaller than the first threshold value α is extracted. Then, four squares around each extracted label position are determined to be invalid areas.

Next, an AND operation is performed between the second label image and the edge extracted image (step 10).
7). FIG. 10 shows a second label image. The line dividing the cell is composed of 10 horizontal lines and 10 vertical lines. The second label image is arranged at the center of each line between adjacent vertical lines in each horizontal line and at the center of each line between adjacent horizontal lines in each vertical line so as to intersect the line. It is an image having rectangular labels B1 to B180.

In the second label image, the labels B1 to B1
Portions other than B180 are black (8-bit binary value “00000000” representing decimal 0). Each label B
The luminance values in 1 to B180 are different 8-bit binary values representing decimal numbers 1 to 180. That is, the luminance value of the label B1 is an 8-bit binary value representing a decimal number 1, the luminance value of the label B2 is an 8-bit binary value representing a decimal number 2, and the luminance value of the label B180 is It is an 8-bit binary value representing the decimal number 180.

When an AND operation is performed between such a second label image and an edge-extracted image, the luminance value becomes 0 in regions other than the labels B1 to B180 in the edge-extracted image. In the area corresponding to each of the labels B1 to B180 in the edge extracted image, in an area where no obstacle such as a hand is present, the luminance value of the line dividing the squares in the area is the brightness value of the corresponding label. The brightness value becomes 0, and the brightness value of a portion other than the line dividing the cells in the area becomes 0.

On the other hand, each label B in the edge extracted image
In the region corresponding to 1 to B180, in the region in which all the lines dividing the cell by the obstacle such as the hand are covered, the luminance values in the region are all 0.

Each label B in the edge extracted image
In the area corresponding to 1 to B180, in the area where a part of the line dividing the cell in the area is covered by an obstacle such as a hand, among the lines dividing the cell in the area,
The luminance value of the part not covered by the obstacle is the luminance value of the corresponding label, and the luminance value of the other parts is zero.

When an AND operation is performed between the second label image and the edge-extracted image, a histogram representing the frequency of the number of strokes with respect to the luminance value is obtained based on the image obtained in step 106, as in step 105. Is created (step 108). Then, second threshold processing is performed based on the created histogram and the second threshold (step 109). That is, a label position corresponding to a luminance value whose histogram value (frequency) is smaller than the second threshold value is extracted. Then, the cell existing at each extracted label position is set as an invalid area.

That is, both the invalid area extracted in step 106 and the invalid area extracted in step 109 are extracted as invalid areas.

The invalid area extracted as described above is
FIG. 11 shows a shaded area. An area other than the invalid area on the shogi board is referred to as an effective area.

According to the above-described invalid area extraction processing, a histogram is created based on the result of the AND operation between the label image and the edge extracted image, and the threshold processing is performed on the histogram. Is extracted,
Processing speed is improved.

It is to be noted that, by using one label image obtained by synthesizing the first label image and the second label image, one AND operation, one histogram creation process, and one threshold process are performed to obtain an invalid area. May be extracted.

Further, even if an invalid area is extracted based on only a label image (first label image) in which a label is set at a position corresponding to each intersection of a vertical line and a horizontal line constituting a line separating the cells, Good.

A label is set at a position corresponding to the center of each line segment between adjacent vertical lines in each horizontal line.
The invalid area may be extracted based only on the label image of A (the image in which only the vertical label is set among the second label images).

A second B label having a label set at a position corresponding to the center of each line segment between adjacent horizontal lines in each vertical line.
(In the second label image, only the horizontally long label is set), the invalid area may be extracted.

Further, an image obtained by combining any two types of label images among the three types of label images of the first label image, the second A label image, and the second B label image may be used as the label image. Good.

FIG. 12 shows a detailed procedure of the frame presence / absence recognition process in step 5 of FIG.

The frame presence / absence recognition processing is performed based on the binarized image of FIG. 6 obtained in step 103 of the invalid area extraction processing. In other words, the edges, ie, the lines dividing the squares on the chess board, the characters drawn on each piece and the outline of the hand are white (8-bit binary value representing 255 in decimal)
11111111 "), and the portion other than the edge portion, that is, the inside of the square on the shogi board and the inside of the outline of the hand are black (8-bit binary value" 00000 representing decimal 0 ").
000 ") is performed based on the edge extracted image.

First, an AND operation is performed between the third label image and the edge extracted image (step 11).
1). FIG. 13 shows a third label image. Third
Is an image having circular labels C1 to C81 at positions corresponding to the center of each square of the shogi board.

In the third label image, the labels C1 to C1
Portions other than C81 are black (8-bit binary value “00000000” representing decimal 0). Each label C1
Luminance values in C81 are 8-bit binary values different from each other and represent decimal numbers 1 to 81. That is, the label C
The luminance value of 1 is an 8-bit binary value representing a decimal number 1, the luminance value of label C2 is an 8-bit binary value of a decimal number 2, and the luminance value of label C81 is a decimal value. It is an 8-bit binary value representing 81. FIG.
Is an enlarged view of the vicinity of the label C1 in the third label image.

When an AND operation is performed between the third label image and the edge extracted image, the luminance value becomes 0 in the area other than the labels C1 to C81 in the edge extracted image. Also, each label C in the edge extraction image
Among the areas corresponding to 1 to C81, the area where no piece exists does not have a luminance value of 0. Therefore,
In the image after the AND operation, among the regions corresponding to the labels C1 to C81 in the edge extracted image, there is no pixel having the luminance value of the label corresponding to the region where no frame exists.

On the other hand, each label C in the edge extracted image
As shown in FIG. 15, in the area corresponding to 1 to C81, in the area where the piece K exists, the brightness value of the character portion of the piece K in the area becomes the brightness value of the corresponding label Ci. , The luminance values of the other parts are 0. Therefore, in the image after the AND operation, among the areas corresponding to the labels C1 to C81 in the edge extracted image, in the area where the piece exists, there is a pixel having the luminance value of the corresponding label Ci. Become.

When the AND operation is performed between the third label image and the edge extracted image in this manner, the frequency of the pixel with respect to the luminance value is represented based on the obtained image as shown in FIG. A histogram is created (step 1
12).

Then, threshold processing is performed based on the created histogram (step 113). That is, it is determined that there is a piece in a square where a label corresponding to a luminance value whose histogram value (frequency) is equal to or larger than the threshold value β is present, and a label corresponding to a luminance value whose histogram value is smaller than the threshold value β is determined. It is determined that there is no piece in the existing square. It should be noted that only the frame presence / absence determination result within the effective area is regarded as the frame presence / absence recognition result.

According to the above-described frame presence / absence recognition processing, a histogram is created based on the result of the AND operation between the label image and the edge extracted image, and threshold value processing is performed on the histogram. Is recognized, so that the processing speed is improved.

FIG. 17 shows the procedure of the interrupt processing. FIG. 18 shows how the state of the queue changes in the interrupt processing.

This interrupt processing is periodically executed based on the timer started in the check processing (step 13).

In the interrupt processing, first, among the ten areas in the queue, the state of the queue in which the input image is to be stored is set to the "image input state" (step 21).
Here, it is assumed that the input images are accumulated in the ten regions in the queue from the lower region. The queue state of each area in the queue includes a state where an image is stored (state with image), a state where no image is stored (state without image), and a state where an image is input (state during image input). And a state in which the image is read and processed (processing state).

As shown in FIG. 18A, when the input image is accumulated in the fifth area from the bottom, the queue state of the sixth area from the bottom is “image input state”. To be.

Next, an image is captured from the CCD camera 1 (step 22). Then, when the image capture from the CCD camera 1 is completed (step 23), FIG.
As shown in (b), the captured image is stored in the area set to the “image input state” (step 24).
Thereafter, the queue state of the area where the image is stored is shown in FIG.
As shown in (c), the state is “image present state”.

FIG. 19 shows the procedure of the image input process in step 3 of FIG. FIG. 20 shows how the state of the queue changes in the image input processing.

In the image input processing, as shown in FIG. 20A, first, the queue state of the area currently in the “processing state” is changed to the “no image state” (step 31). .

Next, it is determined whether or not an area of "image present state" exists in the queue (step 3).
2). If there is no “image present state” area, an image is captured from the CCD camera 1 (step 33). When the timer for interrupt processing is driven by the check processing, the driving of the timer is stopped.

If an area of “image present state” exists, as shown in FIG. 20B, “image present state”
Of the regions where the oldest image is stored, that is, the region with the “image present state”, the queue state of the lowermost region is set to the “processing state”, and then the image is extracted from that region. (Step 34).

FIG. 21 shows the procedure of the check process in step 13 of FIG. FIG. 21 shows the image processing board 3
0 indicates a process.

In the checking process, first, the contents of the queue are cleared (step 41). Next, a timer for starting interrupt processing is started (step 42). Next, a frame arrangement check process is performed (step 43).

In the frame arrangement check processing, first,
The check list created by the host 19 is received (step 51). Checklist for host 1
9 is created as follows. The host 19
As shown in FIG. 22, an arbitrary row or column is randomly selected from the latest recorded game record. And information indicating whether or not there is a piece for each cell on the selected row or column,
If there is a piece, a check list including information indicating the type of the piece is created.

When the checklist is received, as shown in FIG. 22, it is determined whether or not the content of each cell in the row or column corresponding to the checklist in the latest captured image matches the content of the checklist. Is performed (step 52). This determination is performed for each square by performing template matching using a template corresponding to the content of the checklist on the captured image, thereby determining whether the latest captured image matches the content of the checklist. Is done. If the contents of the latest captured image match the contents of the checklist in all the cells (YES in step 53 and YES in step 54), the latest game record stored in the storage device 23 of the host 19 is obtained. Is determined to be correct (OK).

If there is a cell in which the contents of the latest captured image do not match the contents of the checklist (NO in step 53), the latest game record stored in the storage device 23 of the host 19 is erroneously detected. (NG).

When the frame arrangement check processing is completed, the result of the frame arrangement check processing (OKorNG) is transmitted to the host 19.
(Step 44).

If the latest game record stored in the storage device 23 of the host 19 is determined to be correct (OK) in the piece arrangement check processing, the result of step 14 in FIG. Move to In the process of checking the arrangement of the pieces, the storage device 23 of the host 19
Is determined to be incorrect (NG), the latest game record stored in is determined to be NO in step 14 of FIG.
The process proceeds to the recovery process in step S15.

FIG. 23 shows the procedure of the recovery process in step 15 of FIG. FIG. 23 shows the image processing board 3
0 indicates a process.

First, an invalid area extracting process is performed on the currently captured image in the same manner as in step 4 of FIG. 3 (step 61). The invalid area is an area where an obstacle such as a player's hand exists in the captured image.

If it is determined in the invalid area extraction processing that there is no invalid area (obstacle), step 6
Proceed to 4.

In the invalid area extraction processing, when it is determined that there is an invalid area (obstacle) (step 62)
NO), image input processing is performed in the same manner as in step 3 of FIG. 3, and a new image is fetched from the CCD camera 1 or the queue (step 63). Then, the process returns to step 61. Therefore, the processes of steps 61, 62 and 63 are repeated until it is determined that no obstacle exists.

In step 64, the process of acquiring the arrangement of all frames is performed on the captured image determined in step 62 that no obstacle exists.

In the process of acquiring the arrangement of all frames, first, the host 1
9 is received (step 71). This map data is created by the host 19 as follows. As shown in FIG. 24, the host 19 creates map data including information indicating whether there is a piece or not and, if there is a piece, information indicating the type of the piece for each square of the latest recorded game record. .

When the map data is received, as shown in FIG. 24, it is determined whether or not the content of each square of the captured image determined as having no obstacle in step 62 matches the content of the map data. Is determined (Step 7
2). This determination is made for each square by performing template matching using a template corresponding to the content of the map data on the captured image, thereby determining whether the content of the captured image matches the content of the map data.

If the content of the captured image matches the content of the map data (YES in step 73), the position and type of the frame are stored (step 74). Then, it is determined whether or not the processing of step 72 has been completed for all the cells (step 75). If the processing in step 72 has not been completed for all cells, the process returns to step 72, and the processing in step 72 is performed for the next cell.

If the content of the captured image does not match the content of the check list (N in step 73)
O), template matching is sequentially performed on the square image in the captured image using the template of each frame (step 76). If the grid image matches the template image (YES in step 77), the position of the piece and the type of the piece are stored (step 74).

If the cell image does not match the template image (NO in step 77), it is determined whether or not the collation with all types of frames has been completed (step 7).
8). If the collation with all types of pieces has not been completed, the flow returns to step 76, and template matching using the next template is performed.

If it is determined in step 78 that the matching with all types of pieces has been completed, the piece corresponding to the template having the highest correlation value is specified as the piece existing in the corresponding cell. (Step 79). Then, the process proceeds to a step 74, wherein the position of the piece and the type of the piece are stored.

When the processing for acquiring the arrangement of all frames is completed, the information on the position and type of the frames stored in step 74 is transmitted to the host 19 (step 65). Thereafter, the process proceeds to step 10 in FIG. 3, and information on the position and type of the frame transmitted to the host 19 is recorded in the storage device 23 of the host 19.

[0098]

According to the present invention, an apparatus for recognizing the presence or absence of a piece on a shogi board which can improve the processing speed is realized.

[Brief description of the drawings]

FIG. 1 is a schematic diagram showing the appearance of a game record recording device.

FIG. 2 is a block diagram showing an electrical configuration of the game record recording device.

FIG. 3 is a flowchart showing a procedure of a main process of the game record apparatus.

FIG. 4 is a flowchart showing a detailed procedure of an invalid area extraction process in step 4 of FIG. 3;

FIG. 5 is a schematic diagram illustrating an example of a captured image.

FIG. 6 is a schematic diagram showing a binarized image of an image obtained by the edge extraction processing.

FIG. 7 is a schematic diagram showing a first label image.

FIG. 8 shows an AN between a first label image and an edge extracted image.
FIG. 9 is a schematic diagram for explaining an image obtained by a D operation.

FIG. 9 shows an AN between a first label image and an edge extracted image.
9 is a graph showing a histogram created based on an image obtained by a D operation.

FIG. 10 is a schematic diagram showing a second label image.

FIG. 11 is a schematic diagram illustrating an invalid area extracted by an invalid area extraction process.

FIG. 12 is a flowchart showing a detailed procedure of a frame presence / absence recognition process in step 5 of FIG. 3;

FIG. 13 is a schematic diagram showing a third label image.

FIG. 14 is a partially enlarged view of a third label image near a label C1.

FIG. 15 is a diagram showing A between a third label image and an edge extracted image;
It is a schematic diagram for explaining the image obtained by ND operation.

FIG. 16 is a diagram showing A between a third label image and an edge extracted image;
It is a graph which shows the histogram created based on the image obtained by ND operation.

FIG. 17 is a flowchart showing an interrupt process.

FIG. 18 is a schematic diagram illustrating a state in which the state of a queue changes in interrupt processing.

FIG. 19 is a flowchart illustrating a detailed procedure of an image input process in step 3 of FIG. 3;

FIG. 20 is a schematic diagram showing a state in which the state of a queue changes in the image input processing.

FIG. 21 is a flowchart showing a detailed procedure of a check process in step 13 of FIG. 3;

FIG. 22 is a schematic diagram for explaining a check process.

FIG. 23 is a flowchart showing a detailed procedure of a recovery process in step 15 of FIG. 3;

FIG. 24 is a schematic diagram for explaining a recovery process.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 CCD camera 11 Image processing part 13 Image memory 16 CPU 19 Host 23 Storage device 30 Image processing board

Claims (2)

[Claims] 1. Based on a captured image of a piece on a shogi board,
In a piece presence / absence recognition device on a shogi board for recognizing the presence / absence of a piece for each square, a luminance value of a line separating the squares on the shogi board and a luminance value of a character drawn on each piece from a captured image of the piece on the shogi board are first. An edge-extracted image generating means for generating an edge-extracted image which is a luminance value and whose second luminance value is different from the first luminance value of the background inside the squares on the shogi board; Has a label at a position corresponding to the inside of each cell, and a luminance value of a portion other than the label is a third luminance value, and a luminance value in each label is different from the second luminance value and the third luminance value. A predetermined inter-image operation is performed between a label image that is a value and a luminance value that differs for each label, and an edge extraction image, and a first luminance value is calculated in an area corresponding to each label on the edge extraction image. Pixel and the second luminance value Of the pixels, only the pixel of the first luminance value has an image whose luminance value is the luminance value of the corresponding label. Means for creating a histogram representing the frequency of the pixel, and compare the frequency for each obtained luminance value with a threshold value, and if there is a piece in the square where the frequency has a label corresponding to the luminance value equal to or greater than the threshold value A threshold processing means for determining that there is no piece in a cell in which a label corresponding to a luminance value whose frequency is smaller than the threshold value is present, and an invalid area extraction device on a shogi board. . 2. The first luminance value, the second luminance value, the third luminance value, and the luminance value of each label are a predetermined number of bits of 2 bits.
The first luminance value is a binary value in which all bits are represented by 1, and the second luminance value and the third luminance value are binary values in which all bits are represented by 0. The luminance value of each label is set to a different binary value within a range of not more than a binary number in which all bits are 0 and all bits are 1 and all bits are 1. 2. The apparatus according to claim 1, wherein an AND operation is performed between the label image and the edge extracted image.