JP3230371B2 - Character recognition device and character recognition method - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a character recognition device for recognizing characters such as numerals, alphabets and symbols imprinted on an end face of a rolled steel material such as a billet or bloom.

[0002]

2. Description of the Related Art A character recognition device is used, for example, when recognizing a character engraved on an end face of a square pillar-shaped steel material. In such a case, in the conventional character recognition device, first, the end face of the steel material is illuminated by the illuminating means, and a character engraved on the end face is imaged by the CCD camera to obtain a grayscale image. Then, a process of increasing the contrast difference between the character and the background is performed, and thereafter, the grayscale image is converted into a binary image using a threshold value for separating the character and the background. Next, after removing the noise contained in the binary image using a filter,
By comparing the feature amount of the character extracted from the binary image with the corresponding reference feature amount, the character engraved on the steel material is identified.

[0003]

By the way, the steel material is sent to the imaging position by the conveying means. At this time, the steel material is rotated, and the characters are turned sideways or reversed, so that the correct direction is obtained. May not be suitable. The conventional device does not include a mechanism for detecting such rotation. For example, as shown in FIG. 7, when the steel material is rotated by 180 °, the character in the opposite direction is recognized as the character in the correct direction. , There is a problem that characters cannot be recognized. In particular, when the numbers are engraved, and all the numbers are “8” and “6”, when the steel material is rotated by 180 °, the numbers are “8” and “9” by the recognition process. This is problematic because the recognition is incorrect and the result is output as being normal.

The present invention has been made based on the above circumstances, and an object of the present invention is to provide a character recognition device capable of accurately identifying a character even when a test object rotates.

[0005]

Character recognition apparatus according to the invention of claim 1, wherein for achieving the above object, according resolve means for a] is arranged disproportionately on two sides adjacent the square-shaped <br/> like surface The above including characters
X and y directions based on the image obtained by imaging the surface
Find the projection histogram for
The bias of the character arrangement is detected based on the position where the gram is maximum.
And out, a rotation angle detecting means for detecting a rotational angle of the character based on the bias side of the character array, and character segmentation means cuts out the character on the basis of the image, cut out by the character cut-out means And a character recognizing means for identifying the character in consideration of the rotation angle.

[0006]

According to a second aspect of the present invention, in the character recognition apparatus according to the first aspect , the rotation angle detecting means is configured to determine the maximum position and the projection histogram in the vicinity of the maximum position. The bias of the character arrangement is detected based on the distribution status.

[0008]

[0009]

[0010]

[Action] The invention of claim 1, wherein depending on the configuration of the, four
Includes characters that are skewed on two adjacent sides of a square face
By providing the rotation angle detecting means for detecting the deviation of the character arrangement based on the image obtained by imaging the surface, the rotation angle of the character can be detected based on the deviation of the character arrangement. For this reason, since the direction of the character can be known from the rotation angle, even if the character is rotating, the character can be accurately identified in the character identification means in consideration of the rotation angle.

[0011] In the rotation angle detection means, obtains the projection histogram of the x-direction and y-direction, by detecting the deviation of the character array on the basis of the position where the two projection histogram is maximum, the deviation of the character array It can be easily obtained.

According to a second aspect of the present invention, the rotation angle detecting means has a character arrangement based on the position where the projection histogram is maximum and the distribution state of the projection histogram around the position where the projection histogram is maximum. By detecting the deviation, the rotation angle of the character can be detected even if the characters are not arranged on two adjacent sides of the rectangular surface.

[0013]

[0014]

[0015]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS A first embodiment of the present invention will be described below with reference to the drawings. FIG. 1 is a schematic configuration diagram of a character recognition device according to a first embodiment of the present invention, FIG. 2 is a diagram for explaining a method of arranging characters engraved on an end face of a steel material, and FIG. FIG. 4 is a diagram for explaining how to determine the rotation angle of the steel material in the arrangement method, FIG. 4 is a diagram showing an example of a projection histogram in the rotation angle detection unit of the character recognition device in FIG. 1, and FIG. FIG. 5 is a diagram for explaining a method of detecting a rotation angle of a steel material in an angle detection unit.

In the first embodiment, a case is considered in which a square pillar-shaped rolled steel material such as a billet or bloom is used as an object to be inspected, and a character such as a number stamped on an end face of the object to be inspected is identified. These characters are engraved on two adjacent sides out of the four sides of the end face. For example, characters are arranged in a substantially right-angled triangular or substantially L-shaped region, and two sides on both sides of a right-angled portion of this region are located along two adjacent sides on the end face of the inspection object. To place the character. Here, the two sides on both sides of the right angle portion of the area are arranged within a distance of two characters from the two adjacent sides on the end face of the inspection object. In the example shown in FIG. 2A, numbers are arranged in a substantially triangular area, and the first column is
Letters, the second row is three letters, the third row is one letter, and the fourth row is one letter. Then, two sides on both sides of a right angle portion of this region are adjacent to two adjacent sides on the end face of the inspection object.
Also, in FIG. 2B, the alphabet is represented by FIG.
In this example, the inspection object is rotated by 180 °. In the following, as shown in FIG. 2A, a case is considered in which characters are arranged so as to be deviated to the upper left, and in this state, the characters are stamped so as to be in the correct direction. Then, a case where the object to be inspected does not rotate in this state is referred to as a normal state of the object to be inspected.

By arranging the characters in this way, when performing image processing on an image obtained by imaging the object to be inspected, the bias of the character array is detected, and based on the bias of the character array,
The rotation angle of the inspection object can be known. That is, as shown in FIG. 3A, when the character arrangement is shifted to the upper left, it is determined that the inspection object is not rotating, and
As shown in (b), when it is biased to the upper right, 90
° is determined to be rotating, and as shown in FIG.
If it is deviated to the lower right, it is determined that it has rotated 180 °. Then, as shown in FIG. 3 (d), when it is deviated to the lower left, it is determined that it has rotated 270 °.

As shown in FIG. 1, the character recognition apparatus according to the first embodiment includes an illumination unit 12, a CCD camera 14 as an imaging unit, an A / D conversion unit 16, an image memory 18, A processing unit 22, a binary image processing unit 24, a character area cutout unit 26, a rotation angle detection unit 28, a character cutout unit 30, a feature extraction unit 42, and an image recognition processing unit 44 as a character recognition unit. , An image processing memory 46, and a display device 48.

The illumination means 12 is a light source for irradiating the steel material 2. The steel material 2 is transported by the transport means 4 and the illumination means 1
Under the illumination 2, the end face 2 a of the steel material 2 is imaged. The CCD camera 14 forms an optical image of the end face 2a of the steel material 2 as a two-dimensional image on the imaging device by the imaging lens, performs photoelectric conversion on the imaging device, and converts a charge corresponding to the optical image into a one-dimensional image. Output as an electric signal. The one-dimensional electric signal obtained from the image sensor is output to the A / D converter 16 as an image signal. The A / D converter 16 converts an image signal output from the CCD camera 14 into a digital signal and outputs the digital signal. The image memory 18 temporarily stores the image signal (shade image signal) output from the A / D converter 16. Here, the background area where no characters are formed becomes brighter because the light from the illumination means 12 is reflected as it is and the brightness is improved, while the character area is engraved and depressed so that the light is irregularly reflected. The brightness will be lower,
Get dark. Therefore, the density level of the background area is high,
The density level of the character area decreases.

The binarization processing section 22 converts a grayscale image into a binary image based on a predetermined threshold value. Here, a value of “1” is assigned to a pixel (character area) whose density level is equal to or lower than the threshold, and “0” is assigned to a pixel (background area) whose density level is higher than the threshold. The binary image processing unit 24 uses a low-pass filter or the like to remove high-frequency noise included in the binary image, or to perform a process of emphasizing the boundary between the character and the steel material 2. The character area cutout unit 26
The position of the boundary of the steel material 2 is detected, and an area surrounded by the boundary is specified as a character area including a character.

The rotation angle detector 28 determines the rotation angle of the steel material 2 by detecting the deviation of the character arrangement. The character cutout unit 30 specifies the position of each character in the binary image in consideration of the rotation angle of the steel material 2, and has a character string cutout unit 32 and a character line cutout unit 34. The character string cutout unit 32 cuts out each column of the character, and the character line cutout unit 34 cuts out each line of the character.

The feature extracting unit 42 extracts a feature amount of each character cut out by the character cutting unit 30. The image processing memory 46 stores a feature amount serving as a predetermined reference for each character that is known to be engraved in advance. The image recognition processing unit 44 considers the direction of the character based on the rotation angle obtained by the rotation angle detection unit 28, and stores the feature amount extracted from the feature extraction unit 42 with the reference feature amount stored in the image processing memory 46. Characters are identified by comparison. The display device 48 outputs a result of character identification.

Next, the operation of the character recognition device of the first embodiment will be described. First, when the steel material 2 approaches the CCD camera 14, the CCD camera 14 captures an image of the steel material 2 under illumination by the illumination unit 12. The image signal output from the CCD camera 14 is converted into a digital image signal by the A / D converter 16 and then taken into the image memory 18.

The image signal stored in the image memory 18 is
The binarization processing unit 22 converts the image into a binary image based on a predetermined threshold. Then, the binary image is sent to the binary image processing unit 24, and subjected to a low-pass filter or the like to remove high-frequency noise included in the binary image, or to perform a process of emphasizing a boundary between the character and the steel material 2. After that, the character area cutout unit 26 specifies the character area.

Next, the rotation angle detector 28 detects the rotation angle of the steel material 2. First, as shown in FIG. 4, in the character region specified by the character region cutout unit 26, an x-direction projection histogram in which the number of pixels having a value of “1” for each x-coordinate value is calculated, A projection histogram in the y-direction, in which the number of pixels having a value of “1” for each y-coordinate value, is calculated. Then, a point x _{0 at} which the number of pixels is maximum is obtained from the projection histogram in the x direction, and a point y _{0 at} which the number of pixels is maximum is obtained from the projection histogram projected in the y direction. The peak position (x ₀ , y ₀ ) at which the number of pixels thus obtained is the maximum is regarded as representing a bias in the character arrangement.

Next, as shown in FIG. 5, distances L ₁ , L ₂ , L ₃ , and L ₄ from the peak position (x ₀ , y ₀ ) to the four boundaries of the character area are obtained. Here, the distance between the L ₁ distance between the boundary line b ₁ of the peak position and left, L ₂ is the distance between the boundary line b ₂ of the peak position and the right, L ₃ is the peak position and the upper boundary line b ₃ , L ₄ is the distance between the boundary line b ₄ of the peak position and the lower side. Then, the peak position (x ₀ , y
₀ ) is located in the upper left, upper right, lower right, or lower left of the character area to determine the rotation angle of the steel material 2. That is, if L ₁ <L ₂ and L ₃ <L ₄ ,
It can be seen that the peak position is at the upper left of the character area, and it is determined that the steel material 2 is in a normal state (0 ° rotation). L ₁ >
If L ₂ , L ₃ <L _4, it is known that the peak position is at the upper right of the character area, and it is determined that the steel material 2 is rotated by 90 °. If L ₁ > L ₂ and L ₃ > L ₄ , it can be seen that the peak position is at the lower right of the character area,
It is determined that the steel material 2 is rotated by 180 °. Also, L
₁ if a _{<L 2, L 3> L} 4, notice that the peak position in the lower left of the character region, steel 2 is determined to be rotating 270 °.

After the rotation angle of the steel material 2 is detected by the rotation angle detection unit 28, the character string extraction unit 32 and the character line extraction unit 34 extract characters. next,
The feature amount is extracted for each character by the feature extraction unit 42 while considering the direction of the character from the rotation angle of the steel material 2, and then the feature amount is stored in the image processing memory 46 by the image recognition processing unit 44. Characters are identified by comparison with the corresponding reference feature amounts. The result of character identification is output to the display device 48.

In the character recognition device of the first embodiment, when identifying a character engraved on two adjacent sides of the four sides of the end face of the steel material, a rotation angle detecting unit for detecting the deviation of the character arrangement. Is provided, the rotation angle of the steel material can be detected based on how the character arrangement is biased. For this reason, since the direction of the character can be known from the rotation angle,
Even if the steel material is rotating, the characters can be accurately identified.

Next, a character recognition device according to a second embodiment of the present invention will be described. FIG. 6 is a diagram for explaining a method of detecting a rotation angle of a steel material in the character recognition device according to the second embodiment of the present invention. In the second embodiment, similarly to the first embodiment, a square pillar-shaped rolled steel material such as a billet or a bloom is used as an object to be inspected, and characters such as numbers imprinted on an end face of the object to be inspected are identified. Think.

In addition, these characters are stamped at positions biased to two adjacent sides of the four sides of the end face of the steel material, but there are cases where such characters are not stamped at such positions. The character recognition device according to the second embodiment can detect the rotation angle of the steel material even in such a case. This character recognition device has the same configuration as that of the first embodiment, and differs in a method of detecting a rotation angle of a steel material in a rotation angle detection unit. For this reason, in the second embodiment, detailed descriptions of the functions and operations of each unit other than the rotation angle detection unit are omitted.

Next, the operation of the rotation angle detector of the second embodiment will be described. First, similarly to the first embodiment, in the character area, an x-direction projection histogram in which the number of pixels having a value of “1” for each x-coordinate value is calculated, Y that totals the number of pixels having a value of “1”
A projection histogram of the direction is obtained. Then, a point x _{0 at} which the number of pixels is maximum is obtained from the projection histogram in the x direction, and a point y _{0 at} which the number of pixels is maximum is obtained from the projection histogram projected in the y direction. Peak position (x _0, y ₀₎ four boundary b ₁ from the character area, b _2, b _3, a distance L ₁ to b ₄ in which the number of pixels thus obtained is maximum,
L ₂ , L ₃ and L ₄ are obtained.

As shown in FIG. 6, the peak position (x
₀ , y ₀ ), the average value H ₁ ,
H ₂ , H ₃ and H ₄ are obtained. Here, H ₁ is the average value of the projection histogram in the x direction in the range of the width d of the specified size in the −x direction from the peak position, and H ₂ is +
the average value of the projection histogram of the x direction in the range of the width d of the prescribed size in the x direction, H ₃ is the average value of the projection histogram of the y-direction in the range of the width d of the prescribed size in the + y direction from the peak position, H ₄ peak It is the average value of the projection histogram in the y direction in the region of the specified size d in the −y direction from the position. The width d of the specified size is twice as large as the character interval. These average values H ₁ , H ₂ ,
H ₃ and H ₄ represent the distribution status of the projection histogram around the peak position.

Next, it is checked whether or not the character arrangement is biased to two adjacent sides of the steel material. To this end, the distances L ₁ ,
The magnitude relationship between L ₂ , L ₃ , L ₄ and the width d of the specified size is examined. At this time, the following four cases can be considered. case (1) L ₁ or L ₂ <d, L ₃ or L ₄ <d case (2) L ₁ or L ₂ <d, L ₃ and L ₄ > d case (3) L ₁ and L ₂ > d, L ₃ or L ₄ <d case (4) If L ₁ and L ₂ > d, L ₃ and L ₄ > d case (1), the character arrangement is the upper left, upper right, lower right, and lower left of two adjacent sides of the end face of the steel material Is determined to be biased toward any one of. If any of cases (2) to (4),
It is determined that the character array is not biased to two adjacent sides. In particular, case (2) is a case where the character array is vertically shifted, and case (3) is a case where the character array is horizontally shifted. In case (4), the character array is shifted to the center.

The character arrangement is biased on two adjacent sides of the steel material
In the case of Case (1), similarly to the first embodiment, by examining the magnitude relation between L ₁ and L _2, L ₃ and L _4, it detects the rotation angle of the steel material.

When the character arrangement is one of the cases (2) to (4) not biased to two adjacent sides of the steel material, the two sets (L ₁ , L ₂ ), (L ₃ , L ₄ ), for a pair determined to be larger than the width d of the specified size, the magnitude of the average value H _i (i = 1, 2, 3, 4) of the projection histogram having the same subscript as the pair. Examine the relationship. On the other hand, of the two sets (L ₁ , L ₂ ) and (L ₃ , L ₄ ),
For any pair determined to be smaller than the width d of the specified size, the magnitude relationship of L _i (i = 1, 2, 3, 4) of the pair is examined. Then, these magnitude relation, based on the same criteria as for the L _{i (i} = 1,2,3,4) of the first embodiment, it detects the rotation angle of the steel material.

[0036] For example, in the example shown in FIG. 6, L ₁
Since <d, L ₃ and L ₄ > d, it is determined that the character array is shifted downward corresponding to case (2). Next, L
The magnitude relation between ₁ and L ₂ and between H ₃ and H ₄ is examined. In this case, L ₁ <L ₂ and H ₃ <H ₄ . L ₁ <L ₂
, It can be seen that the character arrangement is deviated to the left, and since H ₃ <H ₄ , it can be seen that the characters are distributed more near the lower side from the peak position. Therefore, in the case of FIG. 6, it is determined that the steel material is in a normal state (not rotating).

In the character recognition device of the second embodiment, the rotation of the steel material is determined based on the position (peak position) at which the projection histogram in the x direction and the y direction is maximum, and the distribution of the projection histogram around the peak position. By detecting the angle, the rotation angle of the steel material can be detected even if the character is not biased to any of the upper left, upper right, lower right, and lower left of the end face of the steel material.

The present invention is not limited to the above embodiments, and various modifications can be made within the scope of the invention. For example, in the first and second embodiments, the case where the peak position is obtained by calculating the projection histogram in the x direction and the y direction based on the binary image has been described. A projection histogram in the y direction may be obtained, and the peak position may be obtained from this.

In the first and second embodiments described above, a case where a square pillar-shaped rolled steel material is used as an object to be inspected and characters engraved on an end face of the object to be inspected have been described. When a milk pack or the like is used as the object to be inspected, and the characters engraved on the bottom surface are identified,
The present invention is also applicable to a case where a character printed on a label attached to an end face of a square pillar-shaped test object is identified.

Further, in the first and second embodiments described above, the case where a square pillar is used as the object to be inspected has been described. However, in the present invention, a columnar body having a polygonal cross section is used as the object to be inspected. It is applicable even when used. However, in this case, when a projection histogram is used to determine a peak position, it is necessary to determine a projection histogram for a direction along a predetermined side of a polygonal surface.

[0041]

As described above, according to the first aspect of the present invention, based on an image obtained by capturing an image of a surface including characters arranged on two adjacent sides of a quadrangular surface. The rotation angle of the character array can be detected based on how the character array is biased, so that even if the character is rotated, It is possible to provide a character recognizing device capable of accurately identifying characters in consideration of a rotation angle in the identification means.

[0042] In the rotation angle detection means, obtains the projection histogram of the x-direction and y-direction, by detecting the deviation of the character array on the basis of the position where the two projection histogram is maximum, the deviation of the character array It can be easily obtained.

According to the second aspect of the present invention, the rotation angle detecting means determines the bias of the character arrangement based on the position where the projection histogram is maximum and the distribution state of the projection histogram around the position where the projection histogram is maximum. By performing the detection, it is possible to provide a character recognition device capable of detecting the rotation angle of a character even when the characters are not arranged on two sides adjacent to each other in a rectangular shape.

[0044]

[0045]

[Brief description of the drawings]

FIG. 1 is a schematic configuration diagram of a character recognition device according to a first embodiment of the present invention.

FIG. 2 is a view for explaining a method of arranging characters engraved on an end face of a steel material.

FIG. 3 is a diagram for explaining how to determine a rotation angle of a steel material in the character arrangement method shown in FIG. 2;

FIG. 4 is a diagram illustrating an example of a projection histogram in a rotation angle detection unit of the character recognition device according to the first embodiment.

FIG. 5 is a diagram for explaining a method of detecting a rotation angle of a steel material in the rotation angle detection unit.

FIG. 6 is a diagram for explaining a method of detecting a rotation angle of a steel material in the character recognition device according to the second embodiment of the present invention.

FIG. 7 is a diagram for explaining a normal state in which the steel material is not rotating and a state in which the steel material is rotated by 180 °.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 12 Illumination means 14 CCD camera 16 A / D conversion part 18 Image memory 22 Binary processing part 24 Binary image processing part 26 Character area extraction part 28 Rotation angle detection means 30 Character extraction part 32 Character string extraction part 34 Character line extraction Unit 42 feature extraction unit 44 image recognition unit 46 image processing memory 48 display device

──────────────────────────────────────────────────続 き Continuation of the front page (72) Inventor Satoru Yamaguchi 2-6-3 Otemachi, Chiyoda-ku, Tokyo Nippon Steel Corporation (56) References JP-A-2-247788 (JP, A) Kaihei 2-79184 (JP, A) (58) Field surveyed (Int. Cl. ⁷ , DB name) G06K 9/20 G06T 1/00

Claims (2)

(57) [Claims] 1. A seek projection histogram of the x-direction and y-direction based on the surface containing the character sequence biased toward two sides adjacent the surface of the square shape image obtained by imaging
To the position where the two projection histograms are maximized.
A rotation angle detection unit that detects a bias of the writing character array and detects a rotation angle of the character based on the bias of the character array, a character cutout unit that cuts out the character based on the image, and the character The character cut out by the cutting means,
A character recognition device comprising: a character recognition unit that performs identification in consideration of the rotation angle. 2. The method according to claim 1, wherein the rotation angle detecting means detects the bias of the character arrangement based on the maximum position and a distribution state of the projection histogram around the maximum position. 2. The character recognition device according to 1 .