JPS63223890A - Drawing reader - Google Patents

Abstract

PURPOSE:To recognize alphanumeric characters, KANA (Japanese syllabary) and KANJI (Japanese character) entered into a drawing by providing a means which decides whether character patterns are recognized as alphanumeric characters, or KANA or KANJI individually. CONSTITUTION:Drawing information is inputted through a drawing input part 1 and stored in a frame memory 2 and then a diagram processing process 4 is started to perform a series of linked graphic form processes. Then it is decided whether or not linked areas are character information by checking the sizes of the respective linked areas. When it is decided the linked areas are character information, the processor 4 decides English characters when the shape of a circumscribed frame is rectangular or KANA or KANJI when the shape is square. A character recognizing processor 5 recognizes character patterns cut as a square shape as KANA or KANJI or character patterns cut as a rectangular shape as alphanumeric characters.

Description

DETAILED DESCRIPTION OF THE INVENTION [Object of the Invention] (Industrial Application Field) The present invention relates to a drawing reading device that can effectively read character information written in drawings.

(Prior Art) Drawing reading devices can not only read and recognize information written in drawings, such as logic circuit diagrams and electronic circuit diagrams, but also read and recognize character information written in a space between them. It is necessary to.

Conventionally, the written character information in this type of general drawing is
It is usually alphanumeric only. Since these alphanumeric characters are expressed as a series of character patterns, conventionally the series of patterns are individually cut out from line information other than drawing information that is identified as figures, etc., and these are assumed to be character patterns. Performs character recognition. Therefore, the reading and recognition of character information written in drawings can be achieved through relatively simple processing.

However, with the diversification of drawing information, there is an increasing need to read and recognize drawing information that includes many kana and kanji characters, such as chemical plant drawings.

Furthermore, there is an increasing need to read and recognize drawing information written in a mixture of kana, kanji, and alphanumeric characters.

In order to meet these demands, it is being considered to introduce kana and kanji recognition technology, which has been recently researched and developed. However, this type of conventional character recognition device that recognizes kana and kanji characters is based on the premise that characters written in predetermined squares are recognized for each square. For this reason, it is extremely difficult to detect and recognize characters whose location in a drawing is unknown, that is, characters written at arbitrary positions depending on the content of the drawing.

In particular, unlike the alphanumeric characters mentioned above, kana and kanji are often expressed as a combination of multiple connected patterns, so it is extremely difficult to cut out individual kana and kanji from arbitrary locations in a drawing. There was a problem that.

(Problems to be Solved by the Invention) As described above, the problem with conventional drawing reading devices is that it is difficult to read and recognize alphanumeric characters, kana, and kanji characters arbitrarily written in drawings. was there.

The present invention was made in consideration of these circumstances, and its purpose is to provide a practical and simple structure that can read and recognize alphanumeric characters, kana, and kanji characters mixed in drawings with high precision. An object of the present invention is to provide a drawing reading device with high performance.

[Structure of the Invention] (Means for Solving the Problems) The present invention detects a character writing area from an inputted drawing image, and from this character writing area, for example, forms a series of connected character patterns to form an alphanumeric character. The first character to be cut out and recognized individually as a single character! For example, a set of almost square connected patterns can be written in kana and from the character writing area.
A second character recognition means is provided for cutting out a character pattern of one character forming a kanji and performing recognition processing on each character, and comprehensively judging the recognition results obtained by each of the first and second character recognition means, For example, by employing the recognition results obtained by the first and second character recognition means for the same character pattern, and which have the highest recognition results,
This method is characterized in that recognition results that are consistent with each other are selected as character patterns.

(Operation) According to the present invention, characters written in drawings are recognized by alphanumeric characters and kana and kanji characters. Then, these recognition results are comprehensively judged, and the final recognition result is determined as, for example, a consistent character pattern. That is, if multiple concatenated character patterns exist, each concatenated character pattern is individually recognized as an alphanumeric character, or
It is also checked whether these character pattern sets can be recognized as kana/kanji, and recognition results that are consistent with these connected patterns are selected and adopted.

Therefore, even if alphanumeric characters and kana/kanji characters are mixedly written in the drawing, it is possible to individually recognize these character information with high accuracy.

(Example) Hereinafter, an example of the present invention will be described with reference to the drawings.

FIG. 1 is a schematic configuration diagram of an embodiment of the apparatus. Reference numeral 1 denotes a drawing input unit consisting of a facsimile, drum scanner, hand scanner, TV camera, etc. Through this drawing input unit 1, drawings to be processed are Information is read and input. The drawing information read and input through the drawing input section 1 is stored in the frame memory 2 and used for drawing processing.

The address controller 3 is the frame memory 2
Under this address management, the drawing information stored in the frame memory 2 is read out to the line graphic processing processor 4. This line figure processing processor 4 basically examines the connectivity of, for example, black pixels in the input image (drawing information), and uses the size and length of each connected area to determine if the connected area is a figure, symbol, etc. , or indicates character information. If the connected region is character information, the processing thereof is entrusted to a character recognition processor 5, which will be described later.If the connected region is information such as graphics or symbols, predetermined graphic processing is executed.

Here, the character processor 5 is connected to the line graphics processor 4 via a bus 6, and is also connected to the bus 6.
It is connected to the table processing/control processor 7 via. The table processing control processor 7 controls the graphic processing by the line graphic processor 4 and the character recognition process by the character recognition processor 5 via the bus 6, and also executes graphic processing such as vectorization for line graphics. It is.

FIG. 2 shows the basic flow of processing procedures in this apparatus configured as described above, and the processing functions of this apparatus will be explained according to this processing procedure.

When drawing information is read and input through the drawing input section 1 and stored in the frame memory 2 (step a), the line graphics processing processor 4 is activated. Then, the line figure processing processor 4 executes a series of connected figure detection processes such as checking the connectivity of black pixels in the input image and extracting the connected areas (step b). Then, by checking the size of each connected area, it is determined whether the connected area is character information (step C). At this time, if it is determined that the connected area is character information, the line graphics processing processor 4 detects a circumscribing frame for the character string, and performs processing such as recognizing the inside of the circumscribing frame as a text writing area. are doing.

Then, the data processing/control processor 7 inputs the information of the connected regions determined to be graphic/symbolic information according to the above determination result, and performs predetermined graphic processing such as vectorizing each line segment of each connected figure. (Step d). Then, the figure is recognized from the vectorized line information, and processing such as converting it into a figure code is executed.

In addition, this table processing/control processor 7 inputs the information of the character writing area obtained by the line figure processing processor 4, and detects the character string written therein (
Step e).

This character string detection process takes into consideration the difference between the shape of the circumscribing frame of alphanumeric characters written in the drawing and the shape of the circumscribing frame of kana and kanji characters, and creates a vertically long rectangle as a character pattern consisting of alphanumeric characters. Connected figures each having a circumscribing frame of the shape of a square are cut out individually, and sets of connected figures having a substantially square circumscribing frame are cut out as character patterns forming kana and kanji.

Specifically, when a character pattern as shown in FIG. 3 is given, the line graphic processor 4 divides these connected graphic areas into at, a2, . a3. They are labeled and detected as ~alO. And each of these connected figure areas a1. a2. a3. Information on the circumscribing frame of ~alo, for example, coordinate data (Xa, Ya, Xb.

Yb).

The table processing/control processor 7 converts such information into a table as shown in FIG. Then, the connection relationship of each connected area is investigated from the distance between these connected areas, and if the overall shape of the circumscribed frame of the multiple connected areas with connection ゛ is approximately square,
These are grouped together as a set of connected areas that form kana and kanji. Then, as described above, the information of the connected regions that have been labeled and made into a table are connected using pointers.

On the other hand, even if a character pattern as shown in FIG. 5 is given, the line graphic processor 4 similarly divides these connected graphic areas into bl, b2, . 〜b5 and detected, each of these 3! Figure 1irl area b
1. b2. Information on the circumscribed frame of ~b5 is obtained as coordinate data (Xa, Ya, Xb, Yb), for example.

The table processing/control processor 7 converts such information into a table as shown in FIG. Then, in the same way as in the case of kana and kanji, we examine the connection relationships between these connected areas from the distances between them, and
If the overall circumscribed frame of multiple connected connected areas is a vertically long rectangle, this can be connected to an alphanumeric character wA.
They are grouped into regional groups. Then, as described above, the information of the connected regions that have been labeled and made into a table are connected using pointers. However, when alphanumeric characters are given, in most cases, multiple connected areas are not connected by a pointer, and the character pattern is often expressed using only one connected area.

The table processing/control processor 7 converts the connected areas into a table as described above and connects the connected areas using pointers. However, some alphanumeric characters written in the drawings are written close to each other. In some cases, pseudonyms and
In kanji, hen and labor are sometimes written separately. Therefore, in the table processing/control processor 7, when a connected area or a set of connected areas includes both possibilities of alphanumeric characters or kana/kanji characters, each of them is detected as a character pattern.

Therefore, the character recognition processor 5 has the function of recognizing character patterns cut out with a substantially square circumscribed frame as kana/kanji characters, and the function of recognizing character patterns cut out with a circumscribed frame having a vertically elongated rectangular shape as kana/kanji characters. It has the function of recognizing and processing as

Step f) of individually cutting out sets of connected regions whose circumscribed frames are approximately square, extracting the characteristics of that pattern information, and comparing it with the characteristics of standard patterns for kana and kanji to convert the character pattern into kana and kanji characters. - It is recognized and processed as a kanji (step g).

In addition, in step h), the information of connected regions whose circumscribed frames are vertically elongated rectangular shapes is individually cut out, and the characteristics of the pattern information are extracted and compared with the characteristics of a standard alphanumeric pattern to obtain the character pattern. is subjected to HfJ processing as alphanumeric characters (step i).

After performing the kana/kanji recognition process and the alphanumeric recognition process for the character pattern in this way, the above recognition results for the same character pattern (connected area) are compared with each other in step j). Check for contradictions, etc., and find the recognition result (pseudonym/
The recognition result for the character pattern (recognition result as a kanji character or recognition result as an alphanumeric character) is obtained as a recognition result for the character pattern (step j).

Specifically, as shown in Figure 7, the processing concept is ``When reading and recognizing the kanji ``integrated J,'' the character pattern is first detected.Then, for this character pattern, the kana/kanji are recognized (approximately square) characters are cut out, and each of these character patterns is processed for kana/kanji recognition.

As a result, for the first character, a recognition result of "un" is obtained, and for the second character, a recognition result of "un" is obtained.

On the other hand, when we cut out the characters with alphanumeric characters as the recognition target (the circumscribing frame is a vertical rectangle) for the character pattern "Unified J" mentioned above, the character pattern "Unified" is separated into its parts and labor. Each character is extracted separately and subjected to alphanumeric character recognition. However, since such a standard pattern does not exist for alphanumeric characters, the alphanumeric recognition process fails and "?"? A recognition result of ゛“?゛′ is required.

Therefore, when comparing these recognition results with each other, what is the recognition result for alphanumeric characters? ”, the character pattern is kana/kanji, and the recognition result is “tou” or “go”.
It is determined that

On the other hand, the character pattern ro123J is processed as shown in FIG. In this case, for example, recognition processing is performed on kana and kanji by assuming that "recognition" and "23" are each one character of kana and kanji, and "0" NJ r2J r3J are each an independent alphanumeric character. Recognition processing is performed as follows.

Since the recognition process for kana and kanji fails, a comprehensive determination of both recognition results yields a zero-identification result that the input character pattern is an alphanumeric character rOJ NJ f2J r3J.

For example, if the above-mentioned recognition process for kana/kanji characters and recognition process for alphanumeric characters are performed for the character pattern "hiki", the recognition results "hiki" and r3J MJ may be obtained respectively. . Conversely, for example, if the above-mentioned recognition processing for kana and kanji characters and recognition processing for alphanumeric characters are performed for the character pattern "18j," a recognition result of "old" and a recognition result of rN r8J may be obtained, respectively. .

In such a case, by referring to the context of the character string and the recognition results for other character strings, it is possible to select and determine the one that is consistent with the character string as the correct recognition result. good.

Specifically, recognition results may be selected according to the regularity of the arrangement of character strings.

As described above, this device performs alphanumeric recognition processing and kana-kanji recognition processing for characters written in drawings, and determines whether the recognition results are consistent, or as zero-cognitive results. The final recognition result is one that has good performance and is even more reliable. Therefore, even if kana/kanji and alphanumeric characters coexist in a drawing, each of these characters can be recognized with high accuracy. Moreover, it is possible to effectively read and recognize characters arbitrarily written in a drawing without setting a character entry frame in advance as in conventional character recognition for kana and kanji characters.

Therefore, a great practical effect can be achieved, such as being able to effectively read and input various types of drawing information that are becoming more and more diverse.

Note that the present invention is not limited to the embodiments described above. For example, if the character type can be identified from the connection relationship of connected regions, the shape of its circumscribing frame, etc., execution of recognition processing for another character type may be omitted. Then, the above-described recognition processing for kana/kanji and alphanumeric characters may be performed only for character patterns whose character types are ambiguous. In addition, the present invention can be implemented with various modifications without departing from the gist thereof.

[Effects of the Invention] As explained above, according to the present invention, character patterns are individually cut out by focusing on the difference in the shape characteristics of the circumscribed frames of alphanumeric characters and the shape characteristics of the circumscribed frames of kana and kanji. Against alphanumeric characters! ! ! Perform cognitive processing and recognition processing for kana and kanji, and compare the recognition results to find a consistent Wg.
Since the recognition results are selected, it becomes possible to effectively read and recognize alphanumeric characters, kana, and kanji characters mixed together in the drawing. Moreover, it is possible to read and recognize character information very easily, and it can produce great effects as a highly practical drawing reading device.

[Brief explanation of the drawing]

The figures show one embodiment of the present invention. Fig. 1 is a diagram showing the slope configuration of an actual example device, Fig. 2 is a diagram showing the flow of processing procedures in the embodiment device, and Fig. 3 is a diagram showing the flow of processing procedures in the embodiment device. Figure 4 shows an example of how the kana and kanji information shown in Figure 3 is stored in a table. Figure 5 shows alphanumeric character patterns and their connected areas. Figure 6 is a diagram showing an example of storing the kana/kanji information shown in Figure 5 in a table, and Figures 7 and 8 schematically illustrate examples of recognition processing for character patterns. FIG. 1... Drawing input device, 2... Frame memory, 3
. . . address controller, 4 . . . line graphic processing processor, 5 . . . character recognition processor, 6 . . . bus, 7 . Applicant's agent Patent attorney Takehiko Suzue Figure 1 Figure 3 Figure 5 Character string detection processing

Claims (3)

[Claims] (1) A means for detecting a character writing area in an input drawing image, a first character recognition means for cutting out character patterns as alphanumeric characters from the character writing area and individually recognizing and processing the characters, and the character writing area described above. A second character recognition means extracts character patterns as kana and kanji from the text and recognizes them individually, and the recognition results obtained by the first and second character recognition means are comprehensively evaluated to achieve consistent recognition. 1. A drawing reading device characterized by comprising comprehensive judgment means for selecting a result. (2) The first character recognition means recognizes a series of connected character patterns as one
It is a system that recognizes characters by cutting them out individually as characters.
2. The drawing reading device according to claim 1, wherein the second character recognition means cuts out a set of substantially square connected patterns as one character and recognizes the character. (3) The comprehensive determination means adopts the recognition results obtained by the first and second character recognition means, respectively, for the same character pattern, which have the highest recognition results. Drawing reading device.