JP2929710B2 - Character reading method - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION [Industrial Application Field] The present invention is based on an image obtained by imaging a plurality of unknown characters arranged in a predetermined direction and binarizing them at a predetermined binarization level. By cutting out the pattern of each character and reading the characters from the degree of matching between the pattern and a plurality of preset standard patterns, especially by improving the matching method, the processing time can be reduced and the reading accuracy can be improved. The present invention relates to a character reading method that can be achieved.

[Conventional technology]

Conventionally, as a reading method of this type, matching is performed by aligning the center of a character pattern cut out from a binarized image with the center of each standard pattern, and there is no consistency with any of the standard patterns ( When the matching degree is low), matching is performed while sequentially shifting the positions, and a standard pattern having consistency is extracted. If a standard pattern having a high degree of similarity (large) cannot be extracted by a plurality of matchings, the extracted character is determined to be an unreadable character.

[Problems to be solved by the invention]

As described above, if a similar standard pattern cannot be extracted by the first matching between the cut character pattern and each standard pattern, even when the cut character pattern and the standard pattern are clearly different in shape and size, Until a similar character is detected, matching at different positions is performed a plurality of times, and unnecessary processing is repeated.

In addition, if the density of a character is the same as that of a character to be imaged and the density is different, there is also a problem that the character shape is different in the staircase of binarization, so that it is not possible to read by matching by shifting the position.

Accordingly, it is an object of the present invention to reduce processing time by preventing unnecessary processing, perform appropriate matching processing, and enable highly accurate reading.

[Means for solving the problem]

A plurality of unknown characters arranged in a predetermined direction are imaged, and a pattern of each of the characters is cut out based on an image obtained by binarization at a constant binarization level. In order to read the character from the matching degree with the character pattern, first, the character size in the vertical and horizontal directions is compared between the unknown character pattern and each standard pattern. The first matching process is performed using a difference area value that does not overlap with a standard pattern whose value is within a certain value, and the subsequent matching process is performed for a standard pattern whose difference area value is a certain value or more. Instead of 1) or 2).

1) The second and subsequent matching processes are performed on the remaining standard patterns only by shifting the position of the character in a certain direction (the invention of claim 1).

2) Only the remaining standard patterns are subjected to the second and subsequent matching processes with a plurality of images having different binarization levels (the invention of claim 3).

According to the first aspect of the invention, in performing the second and subsequent matching processes, matching with a plurality of images having different binarization levels can be performed together (the invention of the second aspect).

[Action]

Between the extracted character pattern and each standard pattern,
First, a check is made based on the character size (length and width), and then the matching process is performed to reduce the processing time.

In addition, by performing matching using a plurality of images having different binarization levels, reading can be performed even when the density is different, thereby improving accuracy.

〔Example〕

FIG. 1 is a block diagram showing an embodiment of the present invention.
1 is a television camera, 2 is a binarization circuit, 3 is an image memory, 4 is a memory selection circuit, 5 and 7 are address generation circuits,
6 is a projection detection circuit, 8 is a standard pattern memory, 9 is a matching circuit, 10 is a central processing unit (CPU), and 11 is a bus.

That is, a plurality of binarizing circuits and image memories (CH1 to CHn, M1 to Mn) are provided, respectively. Here, image signals from the television camera 1 are binarized at mutually different levels and stored in the image memory 3. You. The address generation circuit 5 generates an address for scanning the image memory 3 and reads out the contents of the image memory 3 selected by the memory selection circuit 4. Therefore, the projection detection circuit 6 extracts projection data from the contents.

First, a standard character pattern is stored in the image memory 3 via the television camera 1 and the binarization circuit 2, and is stored in the standard pattern memory 8 as a standard pattern.

Next, the unknown character string is stored in the image memory 3 as a binarized image by the television camera 1 and binarization circuit, and projection data is extracted from the image data by the projection detection circuit 6. Further, the CPU 10 extracts individual character patterns based on the projection data.

The pattern thus cut out is checked with all the standard patterns stored in the standard pattern memory 8 based on the character size (length and width), and the following matching process is not performed on the standard pattern having extremely different height and width. If the vertical and horizontal widths are similar to some extent, the center position of the two is adjusted and the difference area is calculated. If the difference is too large, the subsequent matching process is not performed. Is performed, and the result is determined by the CPU 10 to determine consistency. The above points will be specifically described below.

FIG. 2 is an explanatory diagram for explaining an initial frame for character cutout.

Reference numerals 21 to 24 in FIG.
Reference numerals 34 denote initial frames as rectangular frames corresponding to the respective character patterns. That is, the initial frames 31 to 34 are
2424 is set with a certain margin.

FIG. 3 is an explanatory diagram for explaining a character segmentation method.

FIG. 11A shows an intermediate frame obtained by obtaining outer tangents 42 and 43 from both end points of the horizontal (X) projection value 21X of the character pattern in the initial frame 31.
FIG. 11B shows the case where external tangents 52 and 53 are obtained from both end points of the vertical (Y) direction projection value 21Y of the character pattern in the initial frame 31. 43, the case where the circumscribed frame 51 is obtained is shown. Note that the size relationship between these frames is as follows: outer frame <intermediate frame <initial frame.

FIG. 4 is an explanatory diagram for explaining a method of checking a character size (length and width).

This is a comparison of the vertical and horizontal widths of the unknown character pattern A shown in FIG. 1A and the standard pattern B shown in FIG. 2B, where the horizontal width of the unknown character pattern A is LA,
LB is defined as the width of the standard pattern B in the horizontal direction.
If −LB | is equal to or greater than a predetermined value (LMAX), it is assumed that the two are completely different, and no matching process is performed between the unknown character pattern A and the standard pattern B. In other words, only matching with a highly similar standard pattern is performed, and unnecessary matching processing is not performed.
This is to shorten the processing time.

FIG. 5 is an explanatory diagram for describing a first matching processing method between the unknown character pattern A and the standard pattern B.

This is because the centers P1 of the unknown character pattern A and the standard pattern B are overlapped, and the two are overlapped to obtain a non-overlapping portion and the area (difference area) S. As in the case of the above, the two are completely different from each other.
The matching processing after the first time is not performed at all, and unnecessary processing is avoided. It goes without saying that attention may be paid to the area of the overlapping portion instead of the area of the non-overlapping portion.

Thus, the unknown character pattern A and the standard pattern B (B1
-Bn), for those that passed both the vertical and horizontal width (character size) check and the non-overlapping partial area check, perform the second and subsequent matching processes by shifting the center position.

FIG. 6 is an explanatory diagram for explaining an example of such a case.

Here, the unknown character pattern A1 matches any of the standard patterns B1 to B3 in the vertical / horizontal width check, but is selected because the condition check satisfies only the standard pattern B2. Further, in the case where a part of the character is missing like the unknown character pattern A, the center position is shifted and the degree of matching with the standard pattern B2 is low, but the positions of the corners P2 to P5 of both are determined. By performing the matching process in the order, the matching degree can be increased.

By the way, in the case of an unknown character pattern with shading, only a binarized image as shown in FIG. 7 may be obtained at the time of determination with respect to the image at the time of setting as shown in FIG. Therefore, a plurality of levels are set as shown in FIG. 8 (b) in accordance with the degree of shading in CH1 to CHn of the binarization circuit 2 in FIG. 1, and the imaging signal as shown in FIG. Are binarized at these different levels to obtain n kinds of binarized images, which are stored in the image memories M1 to Mn as shown in FIG.
Then, the above-described character extraction and matching processing are performed on each of the binarized images stored in the image memory, and as a result, if there is no consistent standard pattern, the 2D image stored in the image memory is not obtained. The same processing as described above is repeated while sequentially switching the value images. However, image memory
If a consistent standard pattern cannot be extracted up to Mn, the character is regarded as an unreadable character. This processing can be performed in the same manner when the shading varies for each character. Further, the matching process by changing the binarization level may be performed instead of the matching process by shifting the position (the second or later matching process: retry), or may be performed together.

〔The invention's effect〕

According to the present invention, by performing the vertical / horizontal width (character size) check and the difference area check, it is possible to avoid a matching process with a clearly different standard pattern. Processing time can be reduced. In addition, by performing only a check with a different binarization level at the time of retry or in parallel with the matching process of aligning the center position, even if the target character fluctuates due to shading, a more appropriate 2
A matching process can be performed using the binarized image,
High-accuracy reading becomes possible.

[Brief description of the drawings]

FIG. 1 is a block diagram showing a system to which the present invention is applied, FIG. 2 is an explanatory diagram for explaining an initial frame for character extraction, FIG. 3 is an explanatory diagram for explaining a character extracting method, FIG. 4 is an explanatory diagram for explaining a character size checking method, FIG. 5 is an explanatory diagram for explaining a first matching processing method between an unknown character pattern and a standard pattern, and FIG. 6 is a center position. FIG. 7 is an explanatory diagram for explaining the second and subsequent matching processing methods by shifting the position, FIG. 7 is an explanatory diagram for explaining an example of a character pattern having a low density,
FIG. 8 is an explanatory diagram for explaining a binarizing method using different levels, and FIG. 9 is an explanatory diagram for explaining an example of a binarized image binarized at different levels. 1 ... TV camera, 2 ... Binarization circuit, 3 ... Image memory, 4 ... Memory selection circuit, 5,7 ... Address generation circuit, 6 ... Projection detection circuit, 8 ... Standard pattern memory,
9 matching circuit, 10 central processing unit (CPU), 1
1 ... bus, 21-24, 21 '-24', A, A1, A2 ... character pattern, 21X, 21Y ... projection pattern, 31-34 ... initial frame, 41
…… Intermediate frame, 42,43,52,53 …… Circumscribed frame, 51 …… Circumscribed frame,
B, B1 ~ B3 …… Standard pattern.

Claims (3)

(57) [Claims] An image of a plurality of unknown characters arranged in a predetermined direction is taken, and a pattern of each of the characters is cut out based on an image obtained by binarization at a predetermined binarization level. In order to read characters based on the degree of matching with multiple standard patterns, the character size in the vertical and horizontal directions is compared between the unknown character pattern and each standard pattern. Is not performed, a first matching process is performed between the standard pattern having the difference within a certain value and a difference area value that does not overlap with each other. The subsequent matching processing is not performed, and the second and subsequent matching processing is performed by shifting the position of the character in a fixed direction only for the remaining standard patterns. Character reading method characterized by the above-mentioned. 2. The method according to claim 1, wherein the second and subsequent matching processes are performed together with a plurality of images having different binarization levels.
Character reading method described in. 3. A plurality of unknown characters arranged in a predetermined direction are imaged, and a pattern of each of the characters is cut out based on an image obtained by binarization at a predetermined binarization level. In order to read characters based on the degree of matching with multiple standard patterns, first compare the character size in the vertical and horizontal directions between the unknown character pattern and each standard pattern. Without performing the first matching process with a standard pattern whose difference is within a certain value, the first matching process is performed using a difference area value that does not overlap each other. The second and subsequent matching processes are performed for a plurality of images having different binarization levels only for the remaining standard patterns without performing the matching process of And a character reading method.