JPH09305710A - Binarization device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To perform the proper binarization by using a logical difference means to acquire the logical difference image data between the expanded AND image data obtained by an AND means and the contracted binary image data obtained by a binarization means. SOLUTION: An AND device 19 inputs the horizontal and vertical binary image data Dh and Di obtained by the 1st and 2nd binarization devices 16 and 17 respectively and calculates an AND of both data Dh and Di to output the expanded AND image data Dk . A logical difference device 20 calculates and outputs the logical difference image data Dl between the data Dk obtained by the device 19 and the contracted binary image data Dj obtained by a 3rd binarization device 18. The data Dl can acquire the characters having open holes, so that the characters can be easily extracted and recognized at the next stage of the binarization processing. Furthermore, the highly accurate binarization processing is attained for the extraction and recognition of characters.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a binarization device applied to, for example, a pattern recognition device for recognizing a vehicle number or a container number ID.

[0002]

2. Description of the Related Art For example, a vehicle number or a container number ID used in a shipping company is recognized by a pattern recognition device, but the way of writing (format) on the background
Is diversified. For example, the container number ID etc. is written on a corrugated (concave and convex), the container number ID etc. has a character string frame, the number of steps is written over 1 to 3 steps, etc. There is.

As described above, there are various formats for writing the container number ID and the like. However, when the container number ID and the like are imaged by the image pickup device and the characters are recognized, the image data is shaded (black and white). ) Image data.

In the character extraction and character recognition performed on this grayscale (black and white) image data, only the character portions should be highlighted as much as possible and the characters should not be connected.
It is necessary to perform the digitization.

FIG. 15 is a block diagram of such a binarizing device. This binarization device is applied when the characters are brighter than the background, and the target image data D _a obtained by imaging the container number ID or the like is input through the data input device 1 and the target image data D _a is sent to the opening device (hereinafter referred to as an expansion device) 2 and the differential device 3. FIG.
6 (a) shows the luminance cross section of the target image data D _a .

The expansion device 2 has a function of first performing contraction processing and then expansion processing using m × n local regions, and performs contraction / expansion processing of the target image data D _a ,
The opening-processed image data (hereinafter referred to as dilated image data) shown in FIG. 16B is sent to the difference device 3.

The difference device 3 receives the expanded image data and the target image data D _a from the expansion device 2 and outputs expanded difference image data which is the difference between them. The binarization device 4 inputs the expansion difference image data from the difference device 3,
As shown in FIG. 16 (c), the dilation difference image data is binarized by a preset binarization threshold value (fixed threshold value), and the binarized image data shown in FIG. 16 (d) is output. To do.

[0008]

In the binarizing device applied to the image pattern recognizing device for recognizing the vehicle number or the container number ID, the binary morphology is processed by the simple morphological filter process as described above. ing.

However, the vehicle number and the container number I
When the target pattern such as D is written on the corrugated, the brightness of the background part around the character part varies depending on the situation, and the part other than the character part is binarized, and the character extraction after the binarization processing is performed. In the character recognition process, character extraction and character recognition cannot be performed.

Further, when the vehicle number, the container number ID, etc. are surrounded by the character string frame and the characters are written in multiple stages, the gaps in the character string frame are binarized, and as in the above case, 2 It becomes difficult to extract and recognize the character string after the binarization process.

Further, when the space between characters is narrow, there is a case where each character is connected and recognized, which makes character extraction / character recognition difficult. Further, when a character has a hole portion such as the number "6", the hole portion is binarized and it is difficult to recognize the character. Therefore, an object of the present invention is to provide a binarization device capable of highly accurately binarizing a character to be recognized even if the recognition target is written in various formats.

[0012]

According to claim 1, expansion processing means for performing contraction / expansion processing on target image data in at least one of a horizontal direction and a vertical direction; Contraction processing means for performing expansion / contraction processing, and horizontal difference image data and vertical difference image data, which are differences between the expanded image data and target image data in the horizontal and vertical directions obtained by the expansion processing means. Difference means for obtaining contraction difference image data between the contraction image data obtained by the contraction processing means and the target image data, and horizontal difference image data and vertical difference image data obtained by the difference means,
Binarize each contraction difference image data 2
A binarizing unit, a logical product unit that obtains a logical product of the horizontal binary image data and the vertical binary image data that are obtained by the binary unit, and an expansion logic obtained by the logical product unit. A binarizing device comprising: a logical difference means for obtaining a logical difference image data between the product image data and the contracted binary image data obtained by the binarizing means.

In such a binarizing device, the expansion processing means performs contraction / expansion processing on the target image data in at least one of the horizontal direction and the vertical direction to obtain the expanded image data. The contraction processing means performs expansion / contraction processing on the target image data to obtain contracted image data.

Horizontal difference image data and vertical difference image data, which are differences between the expanded image data and the target image data in the horizontal and vertical directions, are obtained, and the contracted difference image between the contracted image data and the target image data is obtained. Ask for data.

Then, these horizontal difference image data,
The vertical difference image data and the contraction difference image data are binarized by the binarizing means, and the logical product of the horizontal direction binary image data and the vertical direction binary image data is logically ANDed. The logical difference image data between the expanded logical product image data and the contracted binary image data calculated by the binarizing means is calculated by the logical difference means.

By such processing, the brightness fluctuation of the background portion due to the corrugation, for example, is eliminated by the contraction / expansion processing in the vertical direction, and proper binarization is performed. When there is a character string frame and the character string extends in multiple stages, the space between the character string frames is prevented from being binarized by the contraction / expansion process in the horizontal direction.

When the gap between characters is small, the characters are prevented from being connected by the expansion / contraction process. The character hole part is not binarized as a hole by expansion / contraction processing.

According to the second aspect, expansion / contraction processing means for performing contraction processing on the target image data in at least one of the horizontal direction and the vertical direction, and contraction / expansion processing for the target image data. Expansion processing means for performing, and horizontal difference image data and vertical difference image data which are differences between the contracted image data obtained by the contraction processing means in the horizontal and vertical directions and the target image data, and the expansion processing means With respect to the difference means for obtaining the expansion difference image data between the expansion image data and the target image data obtained by, and the horizontal direction difference image data, the vertical direction difference image data, and the expansion difference image data obtained by this difference means Binarization means for performing binarization processing, and this two
A logical product means for obtaining a logical product of the horizontal direction binary image data and the vertical direction binary image data obtained by the value converting means, and the contraction logical product image data obtained by the logical product means and the binarization. Logical difference means for obtaining logical difference image data with the expanded binary image data obtained by the means,
It is a binarization device equipped with.

With such a binarization device, the contraction processing means performs expansion / contraction processing in at least one of the horizontal direction and the vertical direction on the target image data, and the expansion processing means performs the target image. The data is subjected to contraction / expansion processing to obtain expansion difference image data.

Horizontal difference image data and vertical difference image data, which are differences between the contracted image data and the target image data in the horizontal and vertical directions, are obtained, and the expansion difference image between the expanded image data and the target image data is obtained. Ask for data.

The horizontal difference image data,
The vertical difference image data and the expanded difference image data are binarized by the binarizing means, and the logical product of the horizontal direction binary image data and the vertical direction binary image data is logically ANDed. The logical difference image data of the contraction logical product image data obtained by the means and the expanded binary image data obtained by the binarizing means is obtained by the logical difference means.

By such processing, similarly to the above, for example, the brightness variation of the background portion due to the corrugated is contracted in the vertical direction.
Eliminate by expansion processing and perform appropriate binarization. When there is a character string frame and the character string extends in multiple stages, the space between the character string frames is prevented from being binarized by the contraction / expansion process in the horizontal direction.

When the space between the characters is small, the characters are prevented from being connected by the expansion / contraction process. The character hole part is not binarized as a hole by expansion / contraction processing.

[0024]

BEST MODE FOR CARRYING OUT THE INVENTION

(1) Hereinafter, a first embodiment of the present invention will be described with reference to the drawings. FIG. 1 is applied to a pattern recognition device 2
It is a block diagram of a quantizer. The target image data D _a obtained by imaging the vehicle number or the container number ID, for example, is input to the data input device 1.

The target image data D _a is, for example, an image of a container number ID written on a corrugated gate as shown in FIG. 2, an image having a character string frame and multiple character strings as shown in FIG. It is an image of a character with a hole like the number "6" or an image of a character string with a narrow character gap like the character "LU".

The data input device 1 includes a first opening device (hereinafter referred to as a first expansion device) 10 and a second opening device (hereinafter referred to as a second expansion device) as expansion processing means. 11, and a horizontal closing device (hereinafter referred to as a horizontal contracting device) 1 as a contraction processing means 1
2 is connected.

Of these, the first expansion device 10 uses the m × 1 local area and performs the contraction process first and then the expansion process in the horizontal direction with respect to the target image data D _a to expand it in the horizontal direction. It has a function of outputting the image data D _b . In addition, m × 1 indicates the kernel size of the filter.

The second expansion device 10 uses the local area of the kernel size 1 × n of the filter, and uses the target image data D _a.
On the other hand, in the vertical direction, a contraction process is performed first, and then an expansion process is performed to output vertically expanded image data D _c .

The horizontal contraction device 12 uses the local area of the kernel size k × l of the filter, and the target image data D
_{It has a} function of performing expansion processing first and then contraction processing in the horizontal direction with respect to a and outputting horizontal contraction image data D _d .

Here, opening processing (contraction-expansion processing)
The closing process (expansion → contraction process) will be described. The definition of expansion is

[0031]

[Equation 1] , The contraction definition is

[0032]

[Equation 2] Is represented by

Wij is an m × n local area centered on the point (i, j), and Wij = {(k, l) | i-[(m-1) / 2] ≤k in the digital image. ≦ i + [m / 2], j − [(n−1) / 2] ≦ l ≦ j + [n / 2]} (3) F (i, j): input image G (i, j): output Therefore, in the expansion process (opening process), contraction is performed first, and then expansion is performed.

[0034]

(Equation 3) In the contraction process (closing process), expansion is performed first, and then contraction is performed.

[0035]

(Equation 4)

For example, when contraction processing by a 3 × 3 local area is performed on the input image data shown in FIG. 4 (a), the contraction processing result becomes the image data shown in FIG. 4 (b). Also, when the input image data shown in FIG. 4 (a) is expanded by a 3 × 3 local area, the expansion result is shown in FIG. 4 (c).
The image data shown in FIG.

On the other hand, when the input image data having the luminance section shown in FIG. 5 (a) is subjected to the expansion processing, the image data of the luminance section shown in FIG. 5 (b) is obtained. When the contraction processing is performed, the image data of the luminance cross section shown in FIG.

Therefore, the image data of this luminance cross section {FIG.
(c)} is a luminance cross section after contraction processing (expansion → contraction) on the input image data. Further, when the contraction process is performed on the input image data shown in FIG. 5 (a), the image data of the luminance cross section shown in FIG. 5 (d) is obtained. When the expansion process is performed on this image data, the same result is obtained. It becomes the image data of the luminance cross section shown in Figure (e).

Therefore, the image data of this luminance cross section {see FIG.
(e)} is the luminance cross section after the expansion processing (contraction → expansion) for the input image data. These first inflator 10,
The second expansion device 11 and the horizontal direction contraction device 12 respectively include first to third difference devices 13 to 1 as difference means.
5 is connected.

The first differential device 13 is the first expansion device 1
The horizontal difference image data D which is the difference between the horizontal direction expanded image data D _b obtained by 0 and the target image data D _a
It has a function to obtain and output _e .

The second differential device 14 is the second expansion device 1
The horizontal difference image data D which is the difference between the vertically expanded image data D _c obtained by 0 and the target image data D _a
It has the function of finding and outputting _f .

The third subtraction device 15 is contraction difference image data D _g which is a difference between the horizontal contraction image data D _d obtained by the horizontal contraction device 12 and the target image data D _a.
It has a function of seeking and outputting.

The first to third difference devices 13 to 15 are respectively connected to first to third binarization devices 16 to 18 as binarizing means. The first and second binarizing devices 16 and 17 are each set with a binarizing threshold value TH _{1 in} advance. Of these, the first binarizing device 16 is 2
Horizontal difference image data D _e by binarization threshold TH ₁
Is binarized to output horizontal binarized image data D _h, and the second binarization device 17 binarizes the vertical difference image data D _f with the binarization threshold TH _1. And has a function of outputting the binary image data D _i in the horizontal direction.

The third binarizing circuit 18 in advance binary threshold value TH ₂ is set, the contraction difference image data D _g The binary threshold value TH ₂ binarization processing to shrink binary It has a function of outputting the _converted image data D _j .

The logical product device 19 calculates the logical product of the horizontal binarized image data D _h and the vertical binarized image data D _i obtained by the first and second binarization devices 16 and 17. It has a function of obtaining the expanded logical product image data D _k .

The logical difference device 20 is a logical difference image between the expanded logical product image data D _k obtained by the logical product device 19 and the contracted binary image data D _j obtained by the third binarizing device 18. It has a function of obtaining and outputting the data D _l .

Next, the operation of the apparatus configured as described above will be described for the case where the container number ID when the character color is lighter than the background color is binarized. (a) A case where the container number ID is written on the corrugate will be described with reference to FIGS. 6 and 7.

The target image data D _a of the container number ID written on the corrugated as shown in FIG. 2 is input to the data input device 1. The target image data D _a has a horizontal luminance cross section as shown in FIG. 6A, for example.

The first expansion device 10 is the data input device 1
If you enter a target image data D _a through, in the horizontal direction with respect to the target image data D _a, first shrinking process is performed, followed expansion processing performed by horizontal expansion image data as shown in FIG. 6 (b) _Db is obtained and sent to the first difference device 13.

The first difference device 13 inputs the horizontally expanded image data D _b obtained by the first expansion device 10 and the target image data D _a through the data input device 1 to obtain the horizontal image data D _a. The horizontal difference image data D _e shown in FIG. 6C, which is the difference between the directional expansion image data D _b and the target image data D _a , is obtained and sent to the first binarization device 16.

[0051] The first binarization unit 16 inputs the horizontal direction differential image data D _e, the binary threshold value TH ₁ relative to the horizontal direction differential image data D _e and binarization Horizontal binarized image data D as shown in FIG.
_{The h} is sent to the AND device 19.

In this horizontal binarized image data D _h ,
As shown in FIG. 6D, a part of the background is binarized in addition to the character part. On the other hand, the target image data D _a is input to the second expansion device 11 via the data input device 1. FIG. 7A shows _a vertical luminance cross section of the target image data D _a .

The second expansion device 11 performs a contraction process first and then an expansion process in the vertical direction with respect to the target image data D _a to obtain a vertical expansion image as shown in FIG. 7B. The data D _c is obtained and sent to the second difference device 14.

The second difference device 14 inputs the vertically expanded image data D _c obtained by the second expansion device 11 and the target image data D _a through the data input device 1 to obtain these vertical image data D _a. The vertical difference image data D _f shown in FIG. 7C, which is the difference between the direction expansion image data D _c and the target image data D _a , is obtained and sent to the second binarization device 17.

[0055] The second binarization unit 17 inputs the vertical difference image data D _f, the binary threshold value TH ₁ relative to the vertical direction differential image data D _f are binarized Vertical binarized image data D as shown in FIG.
_i is sent to the AND device 19.

This logical product device 19 has the first and second two
Horizontal binarized image data D _h and vertical binarized image data D _i obtained by the binarizing devices 16 and 17, respectively.
Is input, the logical product of the horizontal direction binary image data D _h and the vertical direction binary image data D _i is obtained, and the expanded logical product image data D _k is output.

However, this expanded logical product image data D _k
Is generated as image data in which only the character part is binarized. In this way, even if the container number ID is written on the corrugated, it is possible to generate image data in which only the character part is binarized, and to perform character extraction / character recognition processing after this binarization processing. Highly accurate binarization is possible. (b) A case where there is a character string frame and the character string extends in multiple stages will be described with reference to FIGS. 8 and 9.

The target image data D _a of the container number ID having a character string frame as shown in FIG. 3 and having a multi-tiered character string is input to the data input device 1. The target image data D _a has a horizontal luminance cross section as shown in FIG. 8A, for example.

The first expansion device 10 is the data input device 1
When the target image data D _a is input through, the horizontal expansion image data as shown in FIG. 8 (b) is obtained by first performing the contraction process and then the expansion process in the horizontal direction with respect to the target image data D _a . _Db is obtained and sent to the first difference device 13.

The first difference device 13 inputs the horizontally expanded image data D _b obtained by the first expansion device 10 and the target image data D _a through the data input device 1 to obtain the horizontal image data D _a. The horizontal difference image data D _e shown in FIG. 8 (c), which is the difference between the direction expansion image data D _b and the target image data D _a , is obtained and sent to the first binarization device 16.

[0061] The first binarization unit 16 inputs the horizontal direction differential image data D _e, the binary threshold value TH ₁ relative to the horizontal direction differential image data D _e and binarization Horizontal binarized image data D as shown in FIG.
_{The h} is sent to the AND device 19.

In the horizontal binarized image data D _h ,
As shown in FIG. 8D, only the character part is binarized. on the other hand,
The target image data D _a is input to the second expansion device 11 through the data input device 1. FIG. 9A shows the target image data D
shows a vertical luminance section of _a.

The second expansion device 11 first performs contraction processing and then expansion processing in the vertical direction with respect to the target image data D _a , and then performs the vertical expansion image as shown in FIG. 9B. The data D _c is obtained and sent to the second difference device 14.

The second difference device 14 inputs the vertically expanded image data D _c obtained by the second expansion device 11 and the target image data D _a through the data input device 1 to obtain the vertical image data D _a. The vertical difference image data D _f shown in FIG. 9C, which is the difference between the direction expansion image data D _c and the target image data D _a , is obtained and sent to the second binarization device 17.

[0065] The second binarization unit 17 inputs the vertical difference image data D _f, the binary threshold value TH ₁ relative to the vertical direction differential image data D _f are binarized Vertical binarized image data D as shown in FIG.
_i is sent to the AND device 19.

This logical product device 19 has the first and second two
Horizontal binarized image data D _h and vertical binarized image data D _i obtained by the binarizing devices 16 and 17, respectively.
Is input, the logical product of the horizontal direction binary image data D _h and the vertical direction binary image data D _i is obtained, and the expanded logical product image data D _k is output.

However, this expanded logical product image data D _k
Is generated as image data in which only the character part is binarized. Even if there is a character string frame and the character string has multiple stages, it is possible to generate image data in which only the character part is binarized, and to perform character extraction / character recognition processing after this binarization processing. Highly accurate binarization is possible. (c) A case of a character having a hole will be described with reference to FIGS. 10 and 11. Note that, although the horizontal processing is described here, the same applies to the vertical processing.

The target image data D _a of a character with a hole as shown in FIG. 10, for example, the numeral “6” is input to the data input device 1. The target image data D _a has a horizontal luminance cross section as shown in FIG. 11A on the line q ₁ , for example.

The first expansion device 10 is the data input device 1
When the target image data D _a is input through, the horizontal expansion image data as shown in FIG. 11 (b) is obtained by first performing the contraction process and then the expansion process in the horizontal direction with respect to the target image data D _a . _Db is obtained and sent to the first difference device 13.

The first difference device 13 inputs the horizontally expanded image data D _b obtained by the first expansion device 10 and the target image data D _a through the data input device 1 to obtain the horizontal image data D _a. The first binarization device 16 is obtained by obtaining the horizontal difference image data D _e shown in FIG. 11C which is the difference between the direction expansion image data D _b and the target image data D _a .
To send to.

[0071] The first binarization unit 16 inputs the horizontal direction differential image data D _e, the binary threshold value TH ₁ relative to the horizontal direction differential image data D _e and binarization The horizontal binarized image data D _h as shown in FIG. 11D is sent to the AND device 19.

In this horizontal binarized image data D _h ,
As shown in FIG. 11D, the character portion and the hole portion are binarized. On the other hand, the target image data D _a is input to the contraction device 12 through the data input device 1.

The horizontal contraction device 12 first performs expansion processing and then contraction processing on the target image data D _a in the horizontal direction, so that horizontal contraction image data as shown in FIG. Output D _d .

The third difference device 15 is a difference image between the horizontal contraction image data D _d obtained by the horizontal contraction device 12 and the target image data D _{a as} shown in FIG. 11 (f). Data D _g is obtained and output.

[0075] The third binarizing unit 18 inputs the shrinkage difference image data D _g from the third differential unit 15, the binarization threshold value TH ₂ with respect to the contraction difference image data D _g 2 The binarization processing is performed, and the contracted binary image data D _j as shown in FIG. 11 (g) is output.

This contraction binarized image data D _j is shown in FIG.
The hole is binarized as shown in (g). The logical difference device 20 inputs the horizontal binarized image data D _h obtained by the first binarizing device 16 through the logical product device 19 and the contraction obtained by the third binarizing device 18. enter the binary image data D _j, these and horizontal binary image data D _h shrinkage binary image data D _j
The logical difference image data D ₁ shown in FIG. 11 (h) which is the logical difference with

However, as the logical difference image data D _l , it is possible to obtain binarized image data of a character with a hole, for example, the numeral “6”.
The character extraction / character recognition processing in the latter stage of the value conversion processing can be facilitated, and highly accurate binarization can be performed because the character extraction / character recognition processing is performed. (d) A case of a character string having a narrow character spacing will be described with reference to FIGS. 12 and 13. Note that, although the horizontal processing is described here, the same applies to the vertical processing.

The target image data D _{a of a} character string, for example, “LU” having a narrow character spacing as shown in FIG. 12 is input to the data input device 1. The target image data D _a has a horizontal luminance cross section as shown in FIG. 13A on the line q ₂ , for example.

The first expansion device 10 is the data input device 1
When the target image data D _a is input through the horizontal direction image data D _a , the horizontal expansion image data as shown in FIG. Output D _b .

The first difference device 13 is the first expansion device 1
The horizontally expanded image data D _b obtained by 0 is input, and the target image data D is input through the data input device 1.
_{When a} is input, the horizontal difference image data D _e shown in FIG. 13C, which is the difference between the horizontal expansion image data D _b and the target image data D _a , is output.

The first binarization device 16 receives the horizontal difference image data D _e, and inputs the horizontal difference image data D _{e.
The e-} value is binarized by the binarization threshold value TH _{1 and} horizontal binarized image data D _h as shown in FIG.
Is sent to the AND device 19.

In the horizontal binarized image data D _h ,
As shown in FIG. 13 (d), the character part "LU" is also binarized, but the character "L" and the character "U" are also binarized, and these character "L" and "U" Are linked.

On the other hand, the target image data D _a is input to the horizontal contraction device 12 through the data input device 1. The horizontal contraction device 12 first performs expansion processing and then contraction processing on the target image data D _a in the horizontal direction to generate horizontal contraction image data D _d as shown in FIG. Output.

The third difference device 15 is a difference image between the horizontal contraction image data D _d obtained by the horizontal contraction device 12 and the target image data D _{a as} shown in FIG. 13 (f). Data D _g is obtained and output.

[0085] The third binarizing unit 18 inputs the shrinkage difference image data D _g from the third differential unit 15, the binarization threshold value TH ₂ with respect to the contraction difference image data D _g 2 The binarization processing is performed, and the contracted binary image data D _j as shown in FIG. 13 (g) is output.

This contraction binarized image data D _j is shown in FIG.
The character spacing is binarized as shown in (g). The logical difference device 20 inputs the horizontal binarized image data D _h obtained by the first binarizing device 16 through the logical product device 19 and the contraction obtained by the third binarizing device 18. The binarized image data D _j is input, and these horizontal binarized image data D _h and contracted binarized image data D are input.
_The logical difference image data D ₁ shown in FIG. 13 (h), which is the logical difference from _j , is output.

However, even if the logical difference image data D _l is a character string having a narrow character interval, for example, “LU”, the binary image data of the character string having the interval between these characters “L” and “U” is obtained. Therefore, the character extraction / character recognition processing in the latter stage of the binarization processing can be facilitated, and highly accurate binarization can be performed because the character extraction / character recognition processing is performed.

As described above, according to the first embodiment, for example, the brightness variation of the background portion due to the corrugation is eliminated by the contraction / expansion processing in the vertical direction, and the appropriate binarization can be performed. If there is a line and the character string extends in multiple stages, the space between the character string frames cannot be binarized by the contraction / expansion process in the horizontal direction.

Further, when the gap between the characters is small, it is possible to prevent the characters from being connected by the expansion / contraction process, and the character hole portion is not binarized as a hole by the expansion / contraction process.

Therefore, it is possible to perform highly accurate binarization processing on the brightness variation of the background portion due to corrugations, character frames, the number of character string stages, adjacent character strings, hole characters, etc. The extraction / character recognition process can be facilitated. (2) Next, a second embodiment of the present invention will be described with reference to the drawings. The same parts as those in FIG. 1 are denoted by the same reference numerals, and detailed description thereof will be omitted.

FIG. 14 is a block diagram of a binarization device applied to a pattern recognition device. The target image data D _a obtained by imaging the vehicle number or the container number ID, for example, is input to the data input device 1.

The data input device 1 includes a first closing device (hereinafter referred to as a first contracting device) 21 and a second closing device (hereinafter referred to as a second contracting device) as contraction processing means. 22 and a horizontal opening device (hereinafter referred to as a horizontal expansion device) 2 as expansion processing means
3 is connected.

Of these, the first contraction device 21 uses the local area of the kernel size m × 1 of the filter, and first performs the expansion process and then the contraction process in the horizontal direction with respect to the target image data D _a . And has a function of outputting the horizontal contraction image data D _m .

The second contraction device 22 uses the local area of the kernel size 1 × n of the filter, and uses the target image data D _a.
With respect to the vertical direction, it has a function of performing expansion processing first, and then contraction processing to output vertical contraction image data D _n .

The horizontal expansion device 12 uses the local area of the kernel size k × l of the filter to generate the target image data D
_{In the} horizontal direction with respect to a, a contraction process is first performed, and then an expansion process is performed to output the horizontally expanded image data D _r .

First to third difference devices 13 to 15 as difference means are connected to the first contraction device 21, the second contraction device 22 and the horizontal expansion device 23, respectively.

Of these, the first subtraction device 13 is the horizontal contraction image data D _m obtained by the first contraction device 21.
And the target difference image data D _a , the horizontal difference image data D _s is obtained and output.

The second difference device 14 is the second contraction device 2
The horizontal difference image data D which is the difference between the vertical contraction image data D _n and the target image data D _a obtained in Step 2.
_It has the function of calculating and outputting _t .

The third difference device 15 is the expansion difference image data D _u which is the difference between the horizontal expansion image data D _r obtained by the horizontal expansion device 23 and the target image data D _a.
It has a function of seeking and outputting.

The first to third difference devices 13 to 15 are respectively connected to the first to third binarization devices 16 to 18 as binarizing means. The first and second binarizing devices 16 and 17 are each set with a binarizing threshold value TH _{1 in} advance. Of these, the first binarizing device 16 is 2
The horizontal difference image data D _{s according} to the threshold value TH ₁
Is binarized to output horizontal binarized image data D _v, and the second binarization device 17 binarizes the vertical difference image data D _t with the binarized threshold value TH _1. And has a function of outputting the binary image data D _w in the horizontal direction.

In the third binarizing device 18, a binarizing threshold TH ₂ is set in advance, and the dilation difference image data D _u is binarized by the binarizing threshold TH ₂ to dilate the binarization. It has a function of outputting the _converted image data D _x .

The logical product device 19 calculates the logical product of the horizontal binary image data D _v and the vertical binary image data D _w obtained by the first and second binary devices 16 and 17. It has a function of obtaining contraction logical product image data D _y .

The logical difference device 20 calculates the logical difference image between the contracted logical product image data D _y obtained by the logical product device 19 and the expanded binary image data D _x obtained by the third binarizing device 18. It has a function of obtaining and outputting the data D _z .

Next, the operation of the apparatus configured as described above will be described for the case where the container number ID when the character color is lighter than the background color is binarized. (a) The case where the container number ID is written on the corrugate will be described.

The target image data D _a of the container number ID written on the corrugated as shown in FIG. 2 is input to the data input device 1. When the target image data D _a is input through the data input device 1, the first contraction device 21 first performs expansion processing and then contraction processing in the horizontal direction with respect to this target image data D _a . The contracted image data D _m is obtained and sent to the first difference device 13.

The first subtraction device 13 inputs the horizontal contraction image data D _m obtained by the first contraction device 21 and the target image data D _a through the data input device 1 to obtain the horizontal contraction image data D _a. The horizontal difference image data D _s , which is the difference between the direction contraction image data D _m and the target image data D _a , is obtained and sent to the first binarization device 16.

[0107] The first binarization unit 16 inputs the horizontal direction differential image data D _s, the binary threshold value TH ₁ relative to the horizontal direction differential image data D _s and binarization The horizontal binarized image data D _v is sent to the AND device 19.

On the other hand, the target image data D _a is input to the second contraction device 22 through the data input device 1. This second
The contraction device 22 of the first performs the expansion process first in the vertical direction with respect to the target image data D _a , then performs the contraction process to obtain the vertical expansion image data D _n , and sends it to the second difference device 14. .

The second subtraction device 14 inputs the contracted image data D _n in the vertical direction obtained by the second contracting device 22 and the target image data D _a through the data input device 1, The vertical difference image data D _t , which is the difference between the direction expansion image data D _n and the target image data D _a , is obtained and sent to the second binarization device 17.

The second binarization device 17 receives the vertical direction difference image data D _t, and binarizes the vertical direction difference image data D _t with the binarization threshold TH _1. The binary image data D _w in the vertical direction is sent to the AND device 19.

This logical product device 19 has the first and second 2
Horizontal binarized image data D _v and vertical binarized image data D _w obtained by the binarizing devices 16 and 17, respectively.
Is input, the logical product of the horizontal direction binary image data D _v and the vertical direction binary image data D _w is obtained, and the expanded logical product image data D _y is output.

However, this expanded logical product image data D _y
Is generated as image data in which only the character part is binarized. In this way, even if the container number ID is written on the corrugated, it is possible to generate image data in which only the character part is binarized, and to perform character extraction / character recognition processing after this binarization processing. Highly accurate binarization is possible. (b) Explain the case of characters with holes.
Note that, although the horizontal processing is described here, the same applies to the vertical processing.

The target image data D _a of a character having a hole as shown in FIG. 10, for example, the numeral “6” is input to the data input device 1. When the target image data D _a is input through the data input device 1, the first contraction device 21 first performs expansion processing and then contraction processing in the horizontal direction with respect to this target image data D _a . The contracted image data D _m is obtained and sent to the first difference device 13.

The first subtraction device 13 inputs the horizontal contraction image data D _m obtained by the first contraction device 21 and the target image data D _a through the data input device 1 to obtain the horizontal contraction image data D _a. The horizontal difference image data D _s , which is the difference between the direction contraction image data D _m and the target image data D _a , is obtained and sent to the first binarization device 16.

The first binarization device 16 receives the horizontal difference image data D _s, and binarizes the horizontal difference image data D _s with the binarization threshold TH _1. The horizontal binarized image data D _v is sent to the AND device 19.

On the other hand, the target image data D _a is input to the horizontal expansion device 23 through the data input device 1. The horizontal expansion device 23 first performs contraction processing and then expansion processing on the target image data D _a in the horizontal direction, and outputs horizontal expansion image data D _r .

The third difference device 15 is the expansion difference image data D _u which is the difference between the horizontal expansion image data D _r obtained by the horizontal contraction device 12 and the target image data D _a.
Is output.

[0118] The third binarizing unit 18 inputs the expanded difference image data D _u from the third differential unit 15, the binarization threshold value TH ₂ with respect to the expansion difference image data D _u 2 The binarization process is performed, and the expanded binary image data D _x is output.

The logical difference device 20 is the first binarization device 1
6 is input through the AND device 19, and the expanded binary image data D _x obtained by the third binarizing device 18 is input to the horizontal binarized image data D _v. Directional binary image data D _v and dilation 2
The logical difference image data D _z , which is the logical difference from the binarized image data D _x , is output.

However, this logical difference image data D _z
Can obtain binary image data of a character with a hole, for example, the number "6". By opening a hole, character extraction / character recognition processing in the latter stage of the binarization processing can be facilitated. Moreover, since the character extraction / character recognition processing is performed, highly accurate binarization can be performed.

As described above, according to the second embodiment, it is possible to perform highly accurate binarization processing on the brightness variation of the background portion due to corrugations, hole characters, etc. Although the extraction / character recognition process can be facilitated and further explanation is omitted, even if there is a character string frame and the character string extends over multiple levels, or even if the character string has a narrow character spacing, Similar to the form, highly accurate binarization processing can be performed, and character extraction / character recognition processing in the latter stage of the binarization processing can be facilitated.

The present invention is not limited to the first and second embodiments described above, but may be modified as follows. For example, the binarization threshold values TH ₁ and TH ₂ in the _first to third binarization devices 16 to 18 are not limited to fixed values,
It may be a floating threshold.

[0123]

As described in detail above, claim 1 of the present invention,
According to 2, it is possible to provide a binarization device capable of binarizing with high accuracy in order to perform character recognition and the like even if the recognition target is written in various formats.

Further, according to the first and second aspects of the present invention, the brightness variation of the background portion due to the corrugation, the character frame, the number of character string steps,
It is possible to provide a binarization device that can perform highly accurate binarization processing on adjacent character strings, hole characters, etc., and can easily perform character extraction / character recognition processing in the latter stage of the binarization processing.

[Brief description of drawings]

FIG. 1 is a configuration diagram when a first embodiment of a binarizing device according to the present invention is applied to a pattern recognition device.

FIG. 2 is a schematic diagram showing target image data of a container number ID written on a corrugate.

FIG. 3 is a schematic diagram showing target image data having a character string frame and multi-stage character strings.

FIG. 4 is a schematic diagram showing the action of contraction / expansion processing on input image data.

FIG. 5 is a diagram showing the action of contraction / expansion processing seen from the luminance cross section of input image data.

FIG. 6 is a diagram showing a luminance cross-section such as horizontal binarization processing for target image data.

FIG. 7 is a diagram showing a luminance cross section such as binarization processing in the vertical direction with respect to target image data.

FIG. 8 is a diagram showing a luminance cross-section such as horizontal binarization processing for target image data.

FIG. 9 is a diagram showing a luminance cross section such as binarization processing in the vertical direction with respect to target image data.

FIG. 10 is a schematic diagram of target image data of characters with holes.

FIG. 11 is a diagram showing a luminance cross section of a binarization process on target image data.

FIG. 12 is a schematic diagram of target image data of a character string having a narrow character interval.

FIG. 13 is a diagram showing a luminance cross section of a binarization process on target image data.

FIG. 14 is a configuration diagram when a second embodiment of a binarizing device according to the present invention is applied to a pattern recognition device.

FIG. 15 is a block diagram of a conventional binarization device.

FIG. 16 is a diagram showing a binarization process of the apparatus.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 ... Data input device, 10 ... 1st opening device (1st expansion device), 11 ... 2nd opening device (2nd expansion device), 12 ... Horizontal closing device (horizontal contraction device), 13 ... first difference device, 14 ... second difference device, 15 ... third difference device, 16 ... first binarization device, 17 ... second binarization device, 18 ... third binary value Device, 19 ... AND device, 20 ... logical difference device, 21 ... first closing device (first contraction device), 22 ... second closing device (second contraction device), 23 ... horizontal opening Device (horizontal expansion device).

 ─────────────────────────────────────────────────── ─── Continuation of front page (72) Inventor Kyoko Wada 1-1-1 Wadasaki-cho, Hyogo-ku, Kobe-shi, Hyogo Mitsubishi Heavy Industries Ltd. Kobe Shipyard

Claims (2)

[Claims] 1. An expansion processing means for performing contraction / expansion processing on the target image data in at least one of a horizontal direction and a vertical direction, and a contraction processing means for performing expansion / contraction processing on the target image data. A horizontal difference image data and vertical difference image data that are the respective differences between the expanded image data and the target image data in the horizontal and vertical directions obtained by the expansion processing means, and by the contraction processing means. Difference means for respectively obtaining shrinkage difference image data between the obtained shrinkage image data and the target image data, the horizontal direction difference image data, the vertical direction difference image data, and the shrinkage difference image data obtained by this difference means Binarizing means for binarizing each of them, and horizontal binarized image data obtained by the binarizing means. AND means for obtaining a logical product with the binary image data in the vertical direction, and logic for the expanded AND image data obtained by the AND means and the contracted binary image data obtained by the binarizing means A binarization device comprising: a logical difference means for obtaining difference image data. 2. An expansion / contraction processing unit that performs contraction processing on the target image data in at least one of a horizontal direction and a vertical direction, and an expansion processing unit that performs contraction / expansion processing on the target image data. A horizontal difference image data and vertical difference image data, which are the respective differences between the contracted image data and the target image data in the horizontal and vertical directions obtained by the contraction processing unit, and by the expansion processing unit. Difference means for obtaining expansion difference image data between the obtained expansion image data and the target image data, respectively, and the horizontal direction difference image data, the vertical direction difference image data, and the expansion difference image data obtained by this difference means. Binarizing means for binarizing each of them, and horizontal binarized image data obtained by the binarizing means. AND means for obtaining a logical product with the binary image data in the vertical direction, and logic for the contracted AND image data obtained by the AND means and the expanded binary image data obtained by the binarizing means. A binarization device comprising: a logical difference means for obtaining difference image data.