JPH0795332B2 - Character string reader - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION [Object of the Invention] (Field of Industrial Application) The present invention relates to a character string reader for optically reading and recognizing a character string described on a form.

(Prior Art) An optical character reading device (OCR) is a device for reading and inputting a character string written on a form. This kind of device
In general, a character string described in a character entry frame of a predetermined format is optically read as a character image, each character image is recognized, and the information (recognition result) is input.

On the other hand, recently, there is a strong demand that a character reading device also reads and inputs a character string freely (free format) written on a form whose format is not fixed.
However, if the format is not fixed, the reading and recognition of the character string becomes extremely difficult because the orientation of the character string and the orientation of the character are unknown. That is, in the case of Japanese, there are vertical writing and horizontal writing as character string forms. Therefore, when processing an optically read character string image, the orientation of each character that constitutes the character string is determined, and then the orientation of the character is determined and then each character is recognized. Need to be processed. Therefore, there is a problem that not only the processing procedure is complicated but also erroneous recognition and recognition rejection are likely to occur.

(Problems to be solved by the invention) As described above, in the conventional case, in order to read and input a character string from a form in which the character string is described in free form, the direction of each character is checked and then the character is read. Is recognized, the column of the recognition result is comprehensively judged, and the information of the above character string is input.
There was a problem that the processing procedure was complicated.

The present invention has been made in view of such circumstances, and an object thereof is to provide a character string reading device capable of easily and efficiently reading and inputting a character string described in a free format. It is in.

[Structure of the Invention] (Means for Solving Problems) According to the present invention, a photoelectric conversion unit that reads a character string written on a form and a character image from a character string image read by the photoelectric conversion unit are provided. Character detection and extraction means for extracting characters one by one, and the character image extracted by this character detection and extraction means is rotated in 90 ° units with respect to the character reading direction by the photoelectric conversion means, and each character image is processed. To the character rotation recognition means for obtaining the character recognition result for each rotation angle with respect to the character of the recognition result obtained by this character rotation recognition means, by extracting the recognition character of the same rotation angle, the character string for each angle A means for outputting as a candidate, a character string dictionary for storing a character string to be read in advance, and a character string candidate output from the output means for comparison with a read target character string stored in the character string dictionary. And a character string identifying means for obtaining the identification result of the character string by Rukoto.

(Operation) According to the present invention, a character string written on a form is read, character images are sequentially extracted character by character from the read character string image, and the extracted character image is read by the photoelectric conversion means. Rotation is performed in 90 ° units with respect to the direction, and character recognition results are obtained for each rotation angle. Then, for the characters of the obtained recognition result, the character strings candidates for each angle are obtained by extracting the recognized characters of the same rotation angle, and the character string candidates are compared with the character string dictionary stored in advance to determine the characters. The identification result as a column is obtained.

Embodiment An embodiment of the present invention will be described below with reference to the drawings.

FIG. 1 is a schematic configuration diagram of a character string reading apparatus according to an embodiment of the present invention, in which 1 is information on a character string described on a form, which is optically imaged and input as a form image (character string image). It is a photoelectric conversion unit. From the input form image, the character detection unit 2 obtains a projection of the characters in the arrangement direction and a projection in a direction orthogonal to the arrangement direction, as shown in FIG. 2, to obtain the character string image portion and the character string thereof. For example, the character image of each of the constituent characters is detected and cut out in order along the direction of arrangement of the characters. Then, the character detection / cutout unit 2
The individual character images sequentially extracted from the form image are sequentially stored in the pattern memory 3 and used for reading and inputting them.

The character rotation identification unit 4 basically recognizes a character image stored in the pattern memory 3 by referring to the character dictionary 5. At this time, the character rotation identification unit 4 rotates the character image by a predetermined angle with respect to the reading direction of the photoelectric conversion unit 1, and performs rotation processing in units of 90 °, for example, and the orientation is four directions of up, down, left and right. Assuming that there is a possibility that the recognition result can be taken, the recognition result is obtained by referring to the character dictionary 5 for each direction.

Specifically, as illustrated in FIG. When a character string indicating the name "" is given, the first character "" Assuming that the orientation can take four directions, up, down, left and right, the four character images as shown in FIG. 3 obtained by rotating the character image by 90 ° are obtained, and each of these character images is obtained. The recognition process is in progress. By this recognition (identification) process, for example, a character image “0 ° rotated (no rotation)” is displayed. ], The character recognition results such as "Ta" and "Mother" are obtained, and the character images "Yu" and "Shin" are obtained from the 90-degree rotated character image "Yu". Similarly, the character image rotated 180 ° ], The character recognition result such as “T” and “Mother” is obtained, and the character image such as “Kou” and “Use” is obtained from the character image “Kou” rotated by 270 °.

The character string identifying unit 6 compares the character recognition result string obtained for each of the assumed directions as described above with the recognition target character string (for example, name or word) registered in the character string dictionary 7. Collate and the corresponding character string is dictionary 7
It is to determine whether or not it exists inside.

Specifically, as shown in FIG. 4, a sequence of character recognition results identified by being rotated by 0 ° “Tamachi Nozuko” “Mother's veins”…, 90
° Character recognition result sequence identified by rotation "Koji Yura""TamachiShukudan" ..., Character recognition result sequence identified by rotation by 180 ° "Tabata Yankyou", ... etc. The character strings registered in the dictionary 7 as shown in FIG. 5 are collated to find a matching character string. In this case,
The input character string is identified as "Koji Yura" because the corresponding character string is found from the character string dictionary 7 for the string "Koji Yura" of the character recognition result identified by rotating 90 degrees.

It should be noted that the comparison of the character recognition result sequence for each direction and the character string registered in the character string dictionary 7 is performed from top to bottom in the case of vertical writing in Japanese, and in the case of horizontal writing. Under the assumption that the letters are written from left to right in some cases,
The character recognition result sequence rotated by 180 ° and 270 ° is extracted in reverse order and collated with the character string dictionary 7. In other words, for the character recognition result sequence in the direction obtained by rotating 180 °, the dictionary collation is performed as the character string "Keunata Hatada" instead of the character string "Keunata Hata".

As described above, according to the present apparatus, when reading and inputting a character string described in free format, it is assumed that four possible directions of each character forming the character string are up, down, left and right, and the direction is obtained. The obtained character recognition result string is directly collated with the character string dictionary 7 to identify the character string. Therefore, it is not necessary to perform the character string identification processing after determining the character direction one by one as in the conventional case, and the processing procedure can be greatly simplified. Moreover, since the character string is identified and processed as one processing unit, the character string can be accurately recognized, read and input even when the orientation of each character is uncertain. Therefore, the occurrence of erroneous recognition and recognition rejection can be effectively suppressed, and the efficiency of reading and inputting a character string can be improved.

The present invention is not limited to the above embodiment. For example, when extracting individual character images from a character string image, since the number of characters that make up the character string is known, it is possible to use that information to limit the collation range of the character string dictionary 7. is there. Also, the method of recognizing individual character images is not particularly limited. In addition, the present invention can be variously modified and implemented without departing from the scope of the invention.

[Effects of the Invention] As described above, according to the present invention, even if a vertical or horizontal character string written in a free form on a form or the form is vertically inverted, it is very efficient. And it can be read accurately. Therefore, it is possible to greatly simplify the processing as compared with the conventional method that determines the direction of the character and further determines the direction of the character string to identify the character string, and as a result, the recognition time is also shortened. It is possible to obtain a great effect in practical use.

[Brief description of drawings]

FIG. 1 shows an embodiment of the present invention, FIG. 1 is a schematic configuration diagram of a main part of an embodiment device, FIG. 2 is a diagram for explaining an example of an input character string image and a character image extraction process, FIG. 3 is a diagram showing the concept of rotation processing for a character image, FIG. 4 is a diagram showing an example of a sequence of character recognition results obtained for each direction of a character, and FIG. 5 is a recognition target registered in a character string dictionary. It is a figure which shows the example of a character string. DESCRIPTION OF SYMBOLS 1 ... Photoelectric conversion part (photoelectric conversion means), 2 ... Character detection cut-out part (character detection cut-out means), 3 ... Pattern memory, 4 ... Character rotation identification part (character rotation recognition means), 5 ... Character dictionary, 6 …
Character string identifying section (character string identifying means), 7 ... Character string dictionary.

 ─────────────────────────────────────────────────── ─── Continuation of front page (56) Reference JP-A-57-136285 (JP, A) JP-A-58-39377 (JP, A) JP-A-53-29623 (JP, A)

Claims (1)

[Claims] 1. A photoelectric conversion means for reading a character string written on a form, a character detection / cutout means for sequentially extracting character images one by one from a character string image read by the photoelectric conversion means, and the character The character rotation extracted from the detection / cutout means is rotated in units of 90 ° with respect to the character reading direction of the photoelectric conversion means, and character rotation is obtained for each rotation angle for each of the character images. The recognition means and the character of the recognition result obtained by this character rotation recognition means,
A means for outputting a character string candidate for each angle by extracting recognized characters of the same rotation angle, a character string dictionary for storing a character string to be read in advance, and each character string output from the output means. Candidates
A character string reading device, comprising: a character string identifying unit that obtains a result of identifying a character string by comparing it with a read target character string stored in the character string dictionary.