JPS6125284A - Character recognizing device - Google Patents

Abstract

PURPOSE:To improve efficiency of process by searching an untraced stroke on the basis of search specification information, and by discriminating unknown characters from the searched result, and by shortening time for referring to the dictionaries. CONSTITUTION:A character recognizing device is composed of as shown in the figure to repeat and carry out follow-up search of end & point of characters. When search is over and there is not any other end or point of characters, search character information decides whether it is ''1'' or not. If the decision is ''YES'', search is carried out whether or not untraced stroke information, referring to a region of RAM6. In this case, regarding the already searched stroke information converted to eight directions, presence or absence of the untraced stroke information can be easily discriminated because the information is converted into a space code, etc. accordingly. From the result of this discrimination, characters can be discriminated without collating operations.

Description

DETAILED DESCRIPTION OF THE INVENTION Technical Field of the Invention The present invention relates to a character recognition device that automatically reads and recognizes unknown characters such as kana, alphabets, and numbers.

<Background of the Invention> Conventionally, in this type of character recognition device, as shown in the flowchart of FIG. Next, the above binary pattern is pre-processed to make subsequent processing more effective, i.e. noise processing such as black dots on paper P where characters are stored and smoothing of character/figure boundaries. Next, feature extraction processing is performed to extract several features (intersections, branch points, number of loops, stroke length information, etc.) necessary for character recognition. According to this extraction result, some of the characters with common characteristics are narrowed down as candidate characters.This narrowing down completes the first stage of recognition, but if there are multiple candidate characters, the candidate characters are further narrowed down. Detailed identification processing is performed to select the only character from the list.This detailed identification processing is generally referred to as dictionary matching processing.Here, a dictionary is a storage of the characteristics of each character in a fixed series. Usually, it consists of a ROM.Therefore, the above-mentioned narrowing down of candidate characters is called dictionary guidance processing, in which a specific group of characters is selected from a dictionary based on the feature group obtained by feature extraction.

As described above, following this dictionary guidance process, the degree of matching between the standard feature series of each character (hereinafter referred to as standard pattern) stored in advance in the dictionary and the corresponding feature series of the recognition target character is detected. Verification processing is performed. Based on the result of this collation process, the read character is specified, or the recognition result is output as unrecognizable.

Next, an example of a dictionary will be explained.

Figure 7 (Al is a graphic representation of the standard pattern D1 of the kana character "A" in Figure 7 (Bl), and the standard pattern D1 shown here is stored as a dictionary. That is, in this dictionary, the character "A" is ``A'' has three end points ■, one branch point [stock], and 10 directional components (vectors) expressed by number values.

It is made up of. Note that the vector number values shown here are in the stroke direction of the characters from 0 to 7 shown in Figure 8.
This is to distinguish which of the eight directions up to. Therefore, the standard pattern D1 is the character "
A'' starts from the endpoint ■ and has vectors 2, 1, 3, 4, 5 in this order and then ends at the branching point ■, and an element that starts from the endpoint ■ and has vectors 2, 3, 4, 5 and the endpoint ■ In this order, the elements that have the following elements are included.

A diagram illustrating each of these elements is shown in FIG. 7 (lower part of Bl).

Here, a more specific processing procedure of the character recognition device explained using the flowchart of FIG. 6 will be explained using FIG. 7 and FIGS. 9 to 13 described above.

Characters read from paper P are binarized and preprocessed, and then stored in an image memory. The character pattern stored here is a thick one that is the same as the handwriting, as shown in FIG. This thick character pattern is subjected to so-called thinning processing and converted into information indicating the skeleton of the character as shown in FIG. In this conversion, the skeleton is
A (left and right), B (up and down), C (diagonally upward to the right), and D shown in the diagram.
(diagonally downward to the right). The normalization performed here is 9 to facilitate differentiation into the eight directional components shown in FIG. 8. Following this conversion process, a process for obtaining the characteristics of the read character is performed. In this process, a plurality of character features (loops, intersections, branch points, end points, etc.) are extracted, and then dictionary guidance processing is performed using these character features. Once the candidate characters have been narrowed down to a few, the characteristic sequence of the unknown character is then determined and compared with the standard pattern of each candidate character in a dictionary. This feature series is expressed in the same way as the standard pattern D1 in the dictionary, ie, FIG. 7(B), and as a result, the feature series ■134■■34■ as shown in FIG. 12 is obtained. Since this feature series indicates a pattern of the character to be read, that is, an unknown character, it is hereinafter referred to as an unknown pattern.

The unknown pattern obtained as described above is compared with the standard pattern of each candidate character obtained as a result of the dictionary guidance process for its degree of matching. The unknown pattern shown in Figure 12 and the 7th pattern
Assuming that the standard pattern of "A" shown in the figure example is matched, this matching process will be performed as shown in FIG. It is done according to certain rules. In this example, the unknown pattern's directional components are all included in the directional components of the standard pattern and also match in order, so the unknown pattern is recognized and output as a kana character "a".

In a general character recognition device, character recognition is performed as described above, but as can be understood from the above explanation, in the matching process in character recognition, the number of elements making up the unknown pattern and the directional component are It turns out that the order is important.

By the way, the unknown character rA shown in FIGS. 14 and 15
In the case of J and rPJ, when character features of both characters are extracted and compared, each feature such as the number of end points, the number of branch points, and the number of connections are the same. Therefore, regarding these characters, as a result of the dictionary guidance process, both characters rAJ and rPJ are narrowed down as candidate characters. In such a case, in the conventional character recognition device, the unknown pattern and each candidate character rAJrF? The matching process with the Ji standard pattern is performed in a uniform manner, which causes problems such as the dictionary matching operation taking a long time and resulting in a decrease in processing efficiency.

<Object of the Invention> An object of the present invention is to eliminate the drawbacks of the conventional character recognition device described above, shorten the processing time for dictionary comparison, and provide a high-speed and highly accurate character recognition device.

<Configuration and Effects of the Invention> In order to achieve the above object, the present invention stores search designation information that designates whether or not to search for untracked strokes for each character in a dictionary storage means that stores standard patterns. In addition, when determining the feature sequence of an unknown character during a dictionary matching operation, an untracked stroke is searched based on the search designation information, and the unknown character is identified from the search result.

Normally, when finding a feature sequence (unknown pattern) of an unknown character, for example, as shown in Figure 16, scanning for character endpoints is started from the top of the character, and when a character endpoint is detected, stroke tracing is performed from that character endpoint. , character endpoints and branching points to complete the tracking. Thus, for the unknown character rAJ in FIG. 14, the stroke tracing of the part of the character shown in FIG. 17(1) is executed, and for the unknown character rAJ in FIG.
For PJ, stroke tracing of all characters shown in FIGS. 18(1) and 18(2) is executed. Therefore, by implementing the present invention, when searching for unburied strokes for characters "A" and "P" is specified, for unknown characters rAJ, the first
Figure 7 (2) There is an untracked stroke shown in I.
Distinguish between these two characters based on the presence or absence of untracked strokes.
is possible. Therefore, according to the present invention, characters can be identified at a stage before pattern matching begins, the time required for dictionary matching operations can be shortened, and the object of the invention can be achieved, such as contributing to improvement of processing efficiency.

<Description of Embodiments> FIG. 1 shows a character recognition device according to the present invention, in which characters on paper P are optically read by an A/D conversion circuit 1, converted into digital data, and stored in an image memory 2. Ru. -A/
The D conversion circuit 1 includes a CPU 3 and a stroke detection circuit 4.
The CPU 3 executes character recognition processing including matching processing according to the program, and the stroke detection circuit 4 converts read characters into four-directional strokes as shown in FIG. 2, for example.

In the figure, ROM5 is a dictionary and program memory for storing standard patterns of characters to be recognized and search designation information regarding untracked strokes, and RAM5 is a memory for storing 4-direction strokes and 8-direction conversion data of input characters, and a program memory. This is memory for the work area during execution. R
The memory contents of AM6 are stored corresponding to the image after stroke extraction shown in FIG. 2, and the specific contents are shown in FIG. 3.

In FIG. 3, TERM is a storage area for character end point information,
T1 is one end point, SC is the stroke in the C direction, SL
indicates the serial number in the C direction. Similarly, T2 indicates another end point, and SD indicates a stroke in the D direction. Further, R indicates the position of the character endpoint with respect to the stroke, L indicates that the endpoint position is on the left side of the stroke, and K indicates that the endpoint position is on the right side.

CHKN is an area where connection information (reflection points) between strokes is stored, and in the illustrated example, k (right side) of the C direction stroke of serial number S1 and the right side of the D direction stroke of serial number S1 are inflections. It means that they are connected as points. CHBK indicates a storage area for character branch point information, and in the illustrated example, the character branch point B1 is located at the left side of the stroke in the A direction of the serial number S1, and the character branch point B2 is located at the A direction of the serial number S1.
This means that it is located on the k (right side) of the direction stroke.

ASTM is an area where two-dimensional coordinate address information of the end point of the A stroke is stored, and similarly C3TM, DSTM
is an area where two-dimensional coordinate address information of the end points of the C stroke and D stroke is stored. Further, FONT is an area for temporarily storing the tracking process when finding a feature sequence (unknown pattern) of an unknown character, and SCM is an area for storing a feature sequence of the unknown character converted in eight directions.

FIG. 4 shows an example of the dictionary portion of the contents stored in the ROM 5, and the illustrated example shows the character rPJ in FIG. 15. This dictionary stores, for each character storage area, search designation information that designates a search for untracked strokes, followed by a feature series that constitutes a standard pattern. The search designation information is represented by binary data rlJroj, where data "1" means searching for the presence or absence of untracked strokes, while data "0" means searching for the presence or absence of untracked strokes. This means that we do not search for whether or not.

FIG. 5 shows the collation processing operation of the apparatus according to the above embodiment.
Unknown characters - rAJ and character rP when narrowed down to ``
The process of matching J with the standard pattern will be explained.

First, in step 11 (hereinafter referred to as rST11j), a scanning counter is initialized and the image shown in FIG. 2 is scanned on the RAM 5.

In the illustrated example, as shown in FIG. 17, scanning is performed from the left side of the upper side to the right. At this time, the RAM for each scanning point
The two-dimensional address of the stroke end point in the ASTM area DSTM area No. 5 and the coordinates of the scan counter are compared to see if they match (ST12). Unless the coordinates on the comparison result side match, the scan counter is updated and this coordinate address comparison is repeated (sTl2.14.15). If the content of the scan counter and the two-dimensional address of the end point in RAM 6 match, for example, at coordinates (7, 3), this stroke end point is set as a temporary starting point, and then whether or not this stroke end point is a character end point is determined. (STI3). The coordinates (7, 3) of this temporary starting point are stored in the C8TM area, and are assigned to the right end point of stroke number 2 of the C stroke.

However, this end point information does not exist as character end point information in the TER'M area, and therefore, it is determined that the stroke end point -(sc, sl, k) is not a character end point.

As a result, the determination of STI 3 becomes t N 01,
The processing flow of 5T14.15.12 is executed repeatedly. 'Thus, the contents of the scan counter and the two-dimensional address of the end point in RAM5 match at the coordinates (3, 26), but this coordinate is stroke number 1 of the C stroke in the C5TMi area.
Since this stroke end point exists as the middle end point T1 of the TERM area, the end points (sc, sl, L) are determined to be character end points. Then, move the character end mark ■ to 30 in RAM6.
This character end point information (sc, sl,
L) is stored in the FONT area.

By detecting this character end point, 5T13 becomes WYES1, and then in sT16, the conversion process to the 8-direction data shown in FIG. 8 is executed using the stroke information of the FONT area. A stroke from left to right in the C direction stroke in the illustrated example is converted into a direction lO'', and this direction data 10I is stored in the 30M area of the RAM 6.

Next, in order to further search for connection information from the stroke stored in the FONT area, the FONT area is changed to the other stroke end point information (SC, 51, R) of this stroke. And stroke number 1 of C direction stroke
Refer to the data in the FONT area to determine whether the right end (8)) is a character end point or an inflection point.
Examine and check the contents in the TERM and CHKN areas. In the illustrated example, in the CHKN area (SC1S1, R
) can be found, and the determination of "Connection information available" of 5T17 becomes "YES", and the process returns to 5T16, and similarly 8.
A direction conversion process is executed.

In this way, we have reached the stage where the direction data lOl 121 and further the character end point mark ■ have been stored in the 30M area.
The judgment of 17 becomes @N Q # and returns to 5TII,
The scan counter is initialized again and the character endpoint search flow (sTl 2.13.14.15) is repeated. In the case of the unknown character "A" in this embodiment, there are no other character endpoints, so the determination at 5T15 is 'YES' and the process proceeds to 5T18.At this 5T18, whether or not the search designation information is "1" is determined. In this case, the decision is 'YES' (see Figure 4), so ST
In I9, T, E RM area of RAM5, CH
With reference to the KN area and CHBK area, it is searched whether or not untracked stroke information remains. In this case, 8
Since the searched stroke information whose direction has been changed is converted into a space code or the like each time, it is possible to easily determine the presence or absence of untracked stroke information. In the case of this embodiment, there is untracked stroke information in the A direction in the CHBK area, so the determination of 5T20 is "YES", and at this point it can be determined that the unknown character is not "P",
The processing flow in FIG. 17 is completed without the verification operation at T11.

[Brief explanation of drawings]

FIG. 1 is a circuit block diagram of a character recognition device in which the present invention is implemented, FIG. 2 is a diagram showing an example of characters manually entered into the character recognition device shown in FIG. 1, and FIG. 3 is a diagram showing the characters shown in FIG. 1. A diagram showing an example of the storage area arrangement of the RAM of the recognition device, FIG. 4 is the same ROM
5 is a flowchart for explaining the operation of the character recognition device shown in FIG. 1, and FIG. 6 is a general outline of the character recognition device. Flowchart for explanation, FIG. 7 (A)
(Bl is a diagram for explaining a standard pattern example in which the character example "A" is converted into a directional component, Figure 8 is a diagram showing direction numbers when the directional component of a character stroke is normalized into 8 directions, Figure 9 is a diagram showing the optically read pattern of the character "A", Figure 1O is a diagram showing only the stroke skeleton by narrowing the character "A" in Figure 9, and Figure ii1 is 4. A diagram for refining the directional component, Figure 12 is a diagram for explaining the case where the letter "A" shown in Figure 10 is represented by an end point, a directional component, and an inflection point. Figure 13 is the diagram shown in Figure 7. A diagram for explaining the character recognition operation by matching the standard pattern of the character "A" with the characters shown in Figures 10 and 12. Figure 14 is for the character r.
A diagram showing the stroke of AJ, Figure 15 is a diagram showing the stroke of character rPJ, Figure 16 is a diagram explaining the starting point edge for starting the character search scan, and Figures 17 (1) and (2) are diagrams showing the stroke of character rPJ. 1
Figures 18(1) and 18(2) show traced strokes and untracked strokes of the character "A" in Figure 4, and Figures 18(1) and 18(2) depict traced strokes of the character "P". 1... A/D conversion circuit 2... Image memory 3...
CPU 4...Stroke detection circuit 5,...R
OM 6...RAM Otsu Diagram 8 Zuga・0 Diagram Sae 72 Zuzu /2 Erte/3jko

Claims (1)

[Claims] Means for optically reading unknown characters to obtain a character pattern, means for extracting features of the character pattern, means for storing the extracted features, and characters having common extracted features. means for guiding a group as candidate characters; dictionary storage means for storing a unique standard pattern for each character and search designation information for specifying a search for untracked strokes for each character; and the guided character. A character recognition device comprising matching means for searching for features of an unknown pattern for each character of a group based on the search designation information and comparing it with a standard pattern.