KR101070051B1 - recognition method of unchangeability field in dotpattern. - Google Patents

Abstract

The present invention relates to a method for recognizing a constant region in a dot pattern, and more particularly, to a method for recognizing a constant region in a dot pattern for recognizing dot pattern information including a constant region formed in a printed matter.

The method of recognizing the constant region in the dot pattern of the present invention,

A pattern recognition step of recognizing a pattern by the pattern recognition unit when an input image is given;

A gradient value calculating step of analyzing a gradient component having a change in contrast with respect to the input image by a gradient value calculator and classifying the gradient into a binary image;

A blob information acquisition step of acquiring the size and position information (center point) of each blob present in the binarized image by the gradient value calculating step by a blob information acquisition unit;

A clustering step of clustering blobs having values similar to the specific blobs into one by a clustering unit at a specific blob position;

A neighboring blob search step of searching for blobs within a neighboring range for each blob by a neighboring blob search unit;

An invariant region searching step of searching for a list invariant region in the pattern existing in the input image by an invariant region searching unit;

A secondary invariant region searching step of searching for a left invariant region and a right invariant region by a secondary invariant region searching unit based on the list invariant region searched by the invariant region searching step;

A vertical distance calculating step of calculating a vertical distance to be scanned based on the searched invariant region by a vertical distance calculator;

A pattern decoding step of decoding a pattern by a pattern decoding unit having the calculated vertical distance value and the constant region;

A plurality of pattern acquisition steps of acquiring a plurality of patterns obtained by the plurality of pattern acquisition units when a plurality of patterns are obtained from the input image;

A black point discrimination step of determining, by the black point discrimination unit, whether there is a black point within a predetermined range at a specific x, y position;

And a list invariant region searching step of searching by the black point checking unit to determine whether eight consecutive black points have been searched by the black point discrimination step.

According to the present invention, a method of recognizing a constant region in a dot pattern may define a constant region (that is, a fixed region) to determine top, bottom, left, and right sides, and stably handle a case where an incorrect constant region exists (a recognition error or a printing error). Will provide a beneficial effect.

Dot pattern, constant area, information playback.

Description

Recognition method of unchangeability field in dotpattern.}

The present invention relates to a method for recognizing a constant region in a dot pattern, and more particularly, to a method for recognizing a constant region in a dot pattern for recognizing dot pattern information including a constant region formed in a printed matter.

In the past, a voice generating toy for deciphering a specific voice by decoding a barcode printed on a picture book or a game card through an optical sensor has been proposed.

Such a voice generating toy could reproduce various voice information by reading voice information corresponding to the decoded barcode from the memory.

However, such a bar code technology has to secure a dedicated area for bar code printing on the page, and the bar code is for the information processing system to decipher, and it is difficult for the reader to read the picture book or book. Printing bar codes on the page was not only cumbersome for readers, but also reduced the value of books and books.

In addition, due to the practical nature of the terminal for imaging, it is impossible to determine the top, bottom, left, and right of the captured image to define a separate fixed area to determine the top, bottom, left and right, there is an incorrect fixed area (that is, recognition error or printing error) There is a need for a technology that can handle stably for.

An object of the present invention is to recognize the dot pattern information including the constant region formed in the printed matter.

Another object of the present invention is to define a constant area (fixed area) to determine the top, bottom, left and right, and to be able to stably handle a case where there is an incorrect constant area (a recognition error or a printing error).

 Another object of the present invention is to divide and process the invariant region and the available region so that the invariant region serves only to determine whether it is up, down, left, and right dot pattern and to extract available region information (region), In the usable area, a substantially two-dimensional dot pattern is marked to obtain digitized information.

In other words, the constant region is intended to provide the purpose of identifying the location of the available region.

In order to achieve the problem to be solved by the present invention,

Method for recognizing the constant region in the dot pattern according to an embodiment of the present invention,

A pattern recognition step of recognizing a pattern by the pattern recognition unit when an input image is given;

A gradient value calculating step of analyzing a gradient component having a change in contrast with respect to the input image by a gradient value calculator and classifying the gradient into a binary image;

A blob information acquisition step of acquiring the size and position information (center point) of each blob present in the binarized image by the gradient value calculating step by a blob information acquisition unit;

A clustering step of clustering blobs having values similar to the specific blobs into one by a clustering unit at a specific blob position;

A neighboring blob search step of searching for blobs within a neighboring range for each blob by a neighboring blob search unit;

An invariant region searching step of searching for a list invariant region in the pattern existing in the input image by an invariant region searching unit;

A secondary invariant region searching step of searching for a left invariant region and a right invariant region by a secondary invariant region searching unit based on the list invariant region searched by the invariant region searching step;

A vertical distance calculating step of calculating a vertical distance to be scanned based on the searched invariant region by a vertical distance calculator;

A pattern decoding step of decoding a pattern by a pattern decoding unit having the calculated vertical distance value and the constant region;

A plurality of pattern acquisition steps of acquiring a plurality of patterns obtained by the plurality of pattern acquisition units when a plurality of patterns are obtained from the input image;

A black point discrimination step of determining, by the black point discrimination unit, whether there is a black point within a predetermined range at a specific x, y position;

The list invariant region searching step of searching by the black point confirmation unit for searching whether 8 consecutive black points are searched by the black point discrimination step is solved.

As described above, the method of recognizing the constant region in the dot pattern according to the present invention defines a constant region (i.e., a fixed region) to determine the top, bottom, left, and right sides, and is stable in the case of an incorrect constant region (a recognition error or a printing error). It will provide an effect that can be processed by.

In addition, by processing the constant area and the usable area separately, the constant area performs a function of extracting the usable area information (area) by performing only the role of determining whether the dot pattern is up, down, left, right, and correct. The two-dimensional dot pattern is marked to provide an effect of obtaining digitized information.

In addition, even if the positions of the dots are not precise due to a printing error, the recognition is easy, and even when the dots are removed by the printing / recognition error, an error verifiable effect is provided by the continuous dot pattern.

Method for recognizing a constant region in a dot pattern according to an embodiment of the present invention for achieving the above object,

A pattern recognition step of recognizing a pattern by the pattern recognition unit when an input image is given;

A gradient value calculating step of analyzing a gradient component having a change in contrast with respect to the input image by a gradient value calculator and classifying the gradient into a binary image;

A blob information acquisition step of acquiring the size and position information (center point) of each blob present in the binarized image by the gradient value calculating step by a blob information acquisition unit;

A clustering step of clustering blobs having values similar to the specific blobs into one by a clustering unit at a specific blob position;

A neighboring blob search step of searching for blobs within a neighboring range for each blob by a neighboring blob search unit;

An invariant region searching step of searching for a list invariant region in the pattern existing in the input image by an invariant region searching unit;

A secondary invariant region searching step of searching for a left invariant region and a right invariant region by a secondary invariant region searching unit based on the list invariant region searched by the invariant region searching step;

A vertical distance calculating step of calculating a vertical distance to be scanned based on the searched invariant region by a vertical distance calculator;

A pattern decoding step of decoding a pattern by a pattern decoding unit having the calculated vertical distance value and the constant region;

A plurality of pattern acquisition steps of acquiring a plurality of patterns obtained by the plurality of pattern acquisition units when a plurality of patterns are obtained from the input image;

A black point discrimination step of determining, by the black point discrimination unit, whether a black point exists within a predetermined range at a specific x, y position;

And a list invariant region searching step of searching by the black point checking unit to determine whether eight consecutive black points have been searched by the black point discrimination step.

At this time, the pattern is,

An invariant region that is an area that can identify patterns that don't change,

It is characterized by consisting of an available area which is an area capable of generating a pattern.

At this time, the constant region,

A list invariant region consisting of eight consecutive blobs at the top,

Left invariant region consisting of two blobs on the left,

It is characterized by consisting of the right constant region consisting of three blobs on the right.

At this time, the clustering step,

When applied to a large blob, it is characterized by switching to an interactive method to prevent stack overflow.

At this time, the pattern is,

It is formed in a total of 7 x 8 area, except that the 5 x 6 area except the constant area is characterized in that the available area.

At this time, the pattern is,

It is characterized by forming on the printed matter.

At this time, the blob information acquisition step,

Each blob is stored in spatial size as information on the left, top, right, and bottom by the _BLACKDOTREGION structure, and the location of the blob is stored in the center coordinate of the blob by the _POS structure.

Hereinafter, exemplary embodiments of a method of recognizing a constant region in a dot pattern according to the present invention will be described in detail with reference to the accompanying drawings.

1 is a flowchart illustrating a method of recognizing a constant region in a dot pattern according to an exemplary embodiment of the present invention.

2 is an exemplary view showing a pattern example of a method of recognizing a constant region in a dot pattern according to an embodiment of the present invention.

Generally, in order to recognize various multimedia information, a medium such as a printed matter on which a dot pattern in which dots generated according to a dot code generation algorithm are arranged according to a predetermined rule is generated, is decoded by image decoding means as image data, and an image of the corresponding dot pattern portion is generated. The data is coded, and multimedia information corresponding to the coded data is read out from the storage means and played back.

In the dot pattern, multimedia information such as voice means voice, video, video information, or visible, audible, or readable information such as text and symbols. In addition, it can also be referred to as digital data for reproducing a moving picture, voice information, text information, etc. of a computer, a television receiver or a radio broadcasting terminal.

In the dot pattern, code information corresponding to data registered in the storage means is generally defined, XY coordinate values may be defined, and code information and XY coordinate values may be mixed.

The pattern described in the present invention is divided into a constant region and an available region, as shown in FIG.

The constant region is a region defined in order to compensate for rotation and movement when photographed from a pattern recognition apparatus based on a printed pattern on which the pattern is imprinted.

It is a problem to be solved by the present invention to catch the constant region.

The pattern is defined as 5 x 6 free spaces, except for the invariant area of the total 7 x 8 areas, with a value of 1 when black (black point) blobs exist in this area, and a value of 0 when not present. It is defined as having

In addition, considering the number of combinations with values of 0 and 1 in the 5 x 6 region, a pattern of 2 ³⁰ can be generated.

The constant area is used as an area for identifying a pattern that does not change at all times, such as rotation, and the usable area, that is, the usable area is an area within the red border as shown in FIG. It is used as a space to create.

Since the usable area of the present invention has a space of 5 x 6, a pattern of 2 ³⁰ can be generated.

3 is an exemplary view illustrating a schematic concept of a pattern recognition algorithm of a method of recognizing a constant region in a dot pattern according to an embodiment of the present invention.

In other words, when an input image is given, the image goes through DOM Image, AOI Detection, and Pattern Recognition.

In the DOM image, a gradient component having a large change in intensity or grayscale present in the input image is captured.

In principle, the input image contains a lot of noise, so a scale-space that performs noise reduction algorithm on the input image, analyzes the gradient component, or changes the scale value of the Gaussian distribution, In general, the present invention does not implement this process because the CPU does not support the Digital Image Processor, that is, does not have a module optimized for multiplication or division.

In other words, it is assumed that the input image is sufficiently clear by the pattern recognition apparatus without removing the noise.

After DOM image analysis, the input image is classified into a binary image composed of black and white, and the result is shown in FIG. 4.

4 is an exemplary view showing a DOM image example of a method for recognizing a constant region in a dot pattern according to an embodiment of the present invention.

The left picture of FIG. 4 is an input image captured by the pattern recognition device, and the right picture is a picture showing a white pattern on a black background.

Black of FIG. 4 becomes a background, and white becomes characteristic information which forms a pattern.

To extract feature information in binary image, we apply clustering algorithm using connected components.

The clustering algorithm has various methods, but in the present invention, the Connected-Component Analysis algorithm having the lowest computational complexity is used.

This algorithm is also commonly used under the name Grass Labeling algorithm.

The algorithm allocates one clustering pixel having a value equal to its own (in the present invention, a pixel having a white value) in four or eight directions based on the current clustering position.

When the algorithm is repeatedly applied to a pixel having a certain white value present in a binary image, the pixel having all the white values present in the image is labeled as one of specific clusters.

The feature information obtained is defined by the _BLACKDOTREGION structure in the code.

Each blob or cluster stores its spatial size as left, top, right, and bottom information, and the location of the blob is stored in the buffer at the center coordinates of the blob by the _POS structure.

The result of clustering is shown in FIG. 5.

5 is an exemplary view showing an example of AOI Detection of a method of recognizing a constant region in a dot pattern according to an embodiment of the present invention.

The left side of FIG. 5 is a DOM image, and the right side is a photograph showing the result of AOI Detection.

Each of the blobs is drawn with a red rectangle based on the value of _BLACKDOTREGION.

The image is represented by the size information of the blob (by the _BLACKDOTREGION structure) and the center coordinate (by the _POS structure).

The pattern recognition algorithm is applied from the size information and the center coordinate of the image. The key to pattern recognition of the present invention is to find the constant region.

The constant region is divided into an invariant region List, FixedLeft, and FixedRight (see _PATTERN structure) as shown in FIG.

6 is an exemplary diagram illustrating three regions of the constant region of the method of recognizing the constant region in the dot pattern according to an embodiment of the present invention.

The left side is a list invariant region and consists of eight consecutive blobs at the top and is defined as a List in the _PATTERN structure.

The center is a fixedleft constant region, which represents the left constant region and consists of a total of two blobs.

This area is defined as FixedLeft [2] in the _PATTERN structure.

The right side is a FixedRight constant region, which represents the constant region on the right side and consists of a total of three blobs.

This region is defined as FixedRight [3] in the _PATTERN structure.

The pattern recognition algorithm first finds all regions that can be connected to eight consecutive blobs at the top.

Next, it checks whether FixedLeft and FixedRight are detected above and below each candidate area.

Of course, the final candidate obtained through this process is again sampled into the 7 x 8 area and re-validated.

The method of the present invention will be described in detail with reference to FIG. 1.

In the pattern recognition step S110, when the input image is given, the pattern recognition unit recognizes the pattern.

The function for recognizing the pattern is unsigned int PatternRecognition2 (void), as shown in Table 1, and the return value of the function returns the index value of the pattern converted from the available area when the pattern is recognized.

              Table 1

In the gradient value calculating step (S120), the gradient component having a change in contrast is analyzed by the gradient value calculator to classify the input image into a binary image.

The function for analyzing the gradient component is void ComputeDOMImage () as shown in Table 2, and there is no return value of the function.

The main algorithm of the function is to calculate the slope of the input image. In this function, only the slope value of the X axis is obtained.

If the slope value of the X-axis is defined as the coordinate of the current pixel (x, y) and the intensity value corresponding to (x, y) is defined as I (x, y), the X-axis slope is G (x, y). ) = I (x + 1, y) -I (x, y).

When the value is large, the value can be increased positively or negatively.

In this function, G (x, y)> Threshold and G (x, y) <-Threshold are considered as feature points.

The result of the slope is stored in FastBuffer.

[Table 2]

The blob information acquisition step S130 is to acquire the size and position information (center point) of each blob existing in the binarized image by the gradient value calculation step by the blob information acquisition unit.

The function for obtaining the size and position information (center point) of the blob is void clustering (), as shown in Table 3, and there is no return value of the function.

Get each blob present in the image binarized by ComputeDOMImage.

The size and location information (center point) of each extracted blob are obtained and the slope value of the input image is calculated.

Blob size information (left, right, up, down coordinates) is stored in the BlackDotList array (_BLACKDOTREGION structure type), and the center of the blob is stored in the CenterOfBlackDot array (_POS structure).

Each stored size, i.e., the size of the blob and the number of centers, is counted by incrementing by one in nBlackDotPos.

The main algorithm of the function is the ModifiedConnectedComponentLabeling algorithm that implements the Connect-Component algorithm.

In this function, the ModifiedConnectedComponentLabeling algorithm is called repeatedly. Each time it is called, the value of nBlackDotPos is incremented by 1 (one blob is clustered each time), and this value is equal to the Index value of the cluster.

[Table 3]

Clustering step (S140) is to cluster the blobs having a value similar to the particular blob by the clustering unit into one at the position of the particular blob.

As shown in Table 4, the function of clustering into one is void ModifiedConnectedComponentLabeling (int x, int y), and there is no return value of the function.

The function implements the Connected Component Analysis algorithm, and the main algorithm of the function is to bind together a blob having a value similar to itself at the position of the initial blob (in the present invention, at its current position (x of function). , the position given by y), which binds the surrounding white binarized data).

This function is implemented as a recursive call, so there is a risk of stack overflow when a very large blob is applied. Therefore, it is necessary to switch to an iterative method using a stack to prevent this.

Table 4

The adjacent blob search step (S150) is to search for the blobs in the adjacent range for each blob by the adjacent blob search unit.

The function to search for blobs in the adjacent range is void FindNeighborHood (int nSearchMargin), as shown in Table 5. There is no return value of the function.

The reason why the function is required is to eliminate the complexity, since it requires an additional computational amount that always needs to compare the distance between all the blobs in order to find an adjacent area with one blob if the neighbor is not known.

[Table 5]

Invariant region search step (S160) is to search the list invariant region in the pattern existing in the input image by the invariant region search unit.

This function is a function that recognizes the actual pattern. As shown in Table 6, void RangeSearch () has no return value.

This function finds invariant regions defined as List, FixedLeft, and FixedRight as described above.

FindFixedAreaHavingContinuous8BlackDots is a function that finds the list, that is, the top eight consecutive invariant regions at the top, and through this function, if eight consecutive invariant regions are detected, it becomes a candidate region to be patterned. Check if there is any.

Table 6

In the second constant region searching step (S170), the second constant region searching unit searches the left region and the right region on the basis of the list constant region searched by the constant region searching step.

This function is a main function that finds FixedLeft and FixedRight constant regions based on the List constant region. As shown in Table 7, GetPatternWithVariation (int xdiv, int ydiv, int xdiff, int ydiff) has no return value.

The main algorithm of the function calculates the spatial distance to scan vertically from the eight coordinates of the List invariant. This function uses the GetNormalDirection2 function.

The GetPattern3 function reads seven lines and scans them vertically and saves them as a pattern. Then verify the obtained pattern with Verification.

Table 7

In the vertical distance calculation step (S180), the vertical distance calculation unit calculates a vertical distance to be scanned based on the retrieved constant region.

This function calculates the vertical distance to be scanned. As shown in Table 8, void GetNormalDirection2 (LPPattern pPattern, int * pxdiff, int * pydiff, int xdiv, int ydiv) has no return value.

The normal direction means a vertical distance value, and the content of calculating the value forms the content of the function.

Table 8

The pattern decoding step (S190) is to decode the pattern by the pattern decoding unit having the calculated vertical distance value and the constant region.

The function reads a pattern having a vertical distance value and eight constant regions, and as shown in Table 9, void GetPattern3 (LPPattern pArea, int xodiff, int yodiff, int normalx, int normaly, int type). There is no return value.

The function stores the pattern by reading it seven times at a vertical distance based on eight constant regions.

Table 9

In the multi-pattern acquisition step (S200), when a plurality of patterns are obtained from the input image, a plurality of pattern acquisition units acquire a pattern obtained by obtaining the same pattern.

That is, when several patterns are obtained from one image, SINT16 SelectPatternWithPriority () is selected as shown in Table 10 as selecting a pattern in which several identical patterns appear, and the return value of this function is always zero.

Also, the result of this function is that nPatternPos is initialized to 0 if there are two or less identical patterns, that is, only one.

The function's main algorithm is to find the most frequently found pattern value by counting the frequency of the pattern when it comes out.

Table 10

The black point discrimination step S210 determines whether a black point exists within a predetermined range at a specific x and y position by the black point discrimination unit.

That is, as shown in Table 11, SINT16 IsBlackDot (int x, int y, int SEARCHMARGIN) is to check whether there is a black point within a certain SEARCHMARGIN range at a given x and y position, and the return value of the function is out of range. If it returns -1, otherwise, it determines if it is not a black point, and returns 1 if it is a black point, or 0 if it is not.

Table 11

In the list invariant region search step (S220), the black point check unit searches whether eight consecutive black points are searched by the black point discrimination step.

That is, UINT16 FindFixedAreaHavingContinuous8BlackDots (int x, int y, int nCount, SINT16 * pList, int xdiff, int ydiff), as shown in Table 12 as finding eight consecutive points. Returns 0 otherwise.

In other words, we find the top list of constant regions.

Table 12

Through the above steps, a constant area (ie, a fixed area) is defined to determine up, down, left, and right sides, and an error invariant area (a recognition error or a printing error) can be stably provided.

Those skilled in the art to which the present invention pertains as described above may understand that the present invention may be implemented in other specific forms without changing the technical spirit or essential features of the present invention. Therefore, the above-described embodiments are to be understood as illustrative in all respects and not restrictive.

The scope of the present invention is defined by the appended claims rather than the detailed description and all changes or modifications derived from the meaning and scope of the claims and their equivalents are to be construed as being included within the scope of the present invention do.

The method of recognizing the constant region in the dot pattern of the present invention defines a constant region (i.e., a fixed region) to determine the top, bottom, left, and right sides, and can stably handle the case where there is an incorrect constant region (a recognition error or a printing error). To provide a will be widely used in the dot pattern analysis field.

1 is a flowchart illustrating a method of recognizing a constant region in a dot pattern according to an exemplary embodiment of the present invention.

2 is an exemplary view showing a pattern example of a method of recognizing a constant region in a dot pattern according to an embodiment of the present invention.

3 is an exemplary view illustrating a schematic concept of a pattern recognition algorithm of a method of recognizing a constant region in a dot pattern according to an embodiment of the present invention.

4 is an exemplary view showing a DOM image example of a method for recognizing a constant region in a dot pattern according to an embodiment of the present invention.

5 is an exemplary view showing an example of AOI Detection of a method of recognizing a constant region in a dot pattern according to an embodiment of the present invention.

6 is an exemplary diagram illustrating three regions of the constant region of the method of recognizing the constant region in the dot pattern according to an embodiment of the present invention.

Claims (7)

A pattern recognition step of recognizing a pattern by the pattern recognition unit when an input image is given; A gradient value calculating step of analyzing a gradient component having a change in contrast with respect to the input image by a gradient value calculator and classifying the gradient into a binary image; A blob information acquisition step of acquiring the size and position information (center point) of each blob present in the binarized image by the gradient value calculating step by a blob information acquisition unit; A clustering step of clustering blobs having values similar to the specific blobs into one by a clustering unit at a specific blob position; A neighboring blob search step of searching for blobs within a neighboring range for each blob by a neighboring blob search unit; An invariant region searching step of searching for a list invariant region in the pattern existing in the input image by an invariant region searching unit; A secondary invariant region searching step of searching for a left invariant region and a right invariant region by a secondary invariant region searching unit based on the list invariant region searched by the invariant region searching step; A vertical distance calculating step of calculating a vertical distance to be scanned based on the searched invariant region by a vertical distance calculator; A pattern decoding step of decoding a pattern by a pattern decoding unit having the calculated vertical distance value and the constant region; A plurality of pattern acquisition steps of acquiring a plurality of patterns obtained by the plurality of pattern acquisition units when a plurality of patterns are obtained from the input image; A black point discrimination step of determining, by the black point discrimination unit, whether there is a black point within a predetermined range at a specific x, y position; And a list invariant region searching step of searching by the black point confirmation unit for whether eight consecutive black points are searched by the black point discrimination step. The method of claim 1, The pattern is, An invariant region that is an area that can identify patterns that don't change, A method for recognizing an invariant region in a dot pattern, characterized in that it consists of an available region which is an area capable of generating a pattern. 3. The method of claim 2, The constant region is, A list invariant region consisting of eight consecutive blobs at the top, Left invariant region consisting of two blobs on the left, A method for recognizing a constant region in a dot pattern, characterized in that the right constant region consists of three blobs on the right side. The method of claim 1, The clustering step, A method for recognizing invariant regions in a dot pattern, characterized in that when applied to a large blob, to switch in an interactive manner to prevent stack overflow. The method of claim 1, The pattern is, A method of recognizing a constant region in a dot pattern, which is formed in a total of 7 x 8 regions, except that the 5 x 6 region except for the constant region is an available region. The method of claim 1, The pattern is, A method for recognizing a constant region in a dot pattern, characterized in that formed on the print. delete

Images