JPS6343789B2 - - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention particularly relates to a character reading device that can effectively extract and read each character pattern from a character string pattern with a known average pitch.

Recently, various character reading devices including OCR have been developed and are greatly contributing to business processing.
Generally, a character string can be viewed as a group of characters written in character frames arranged at a predetermined pitch, and when this is read and recognized, character detection and extraction processing is performed. Accurately cutting out this character frame is very important in recognition processing, and conventionally this has been done, for example, as follows.

For example, a mark indicating the character position is attached to a medium such as a form in advance, and this is detected to detect and cut out the character, or the expansion of the character itself is detected to confirm the character position and then the character is detected and cut out. are doing.
However, in the former method, the presence of extra marks on the medium makes handling of the medium, such as forms, complicated, and when the marks are small, incorrect position recognition may occur due to the influence of noise due to adhesion of dust, etc. There was also some fear. Furthermore, eliminating the influence of the noise requires a fairly complicated device mechanism, which poses a problem of poor practicality. On the other hand, in the latter method,
For characters such as "1" and "I" that are not written to fill the entire character frame, there will be many errors in the detected extraction position, resulting in a so-called single-cut state with respect to other characters. I was afraid. Furthermore, since the position is detected from the character itself to be read, the device configuration and detection process are not only complicated, but also there are major problems such as vulnerability to noise.

The present invention was made in consideration of these circumstances, and its purpose is to provide a practical method that can easily, stably, and accurately cut out character strings with known average pitches. The objective is to provide a high quality character reading device.

Embodiments of the present invention will be described below with reference to the drawings.

Figure 1 shows an example of a character string pattern with a known average arrangement pitch. Each character 1 is assumed to be written or printed within a character frame 2 arranged at a predetermined pitch. I can do it. As illustrated in this figure, character 1 is "I"
There is a mixture of words such as ``W'' written only in the center of frame 2 and words such as ``W'' that extend over the entire area of frame 2.

Therefore, for such a character string pattern, a cutting reference position is first temporarily set, and the character string pattern is cut out using the same position as a reference at the above-mentioned average arrangement pitch as shown by the dashed line in the figure. will be held. Each of the patterns cut out in this way is subjected to superimposition processing as shown in FIG. 2A as the next step, and its projected component is determined as shown in FIG. 2B. This superimposition process is performed by, for example, 5
It is carried out over 6 patterns. However, the projection components of the overlapping pattern appear corresponding to the cutout positions, and the state in which the character 1 is divided into two parts is often concentrated on both sides of the overlapping pattern. Then, a blank portion corresponding to the space between the character frames 2 exists in the middle position between both sides where the amount of information of these projected components is large. Incidentally, if the tentatively set cutout reference position matches the frame 2, a projected component will appear in the center of the overlapping pattern, and blank areas will appear on both sides. Note that this blank area is detected by, for example, discriminating projected components at a predetermined level and recognizing components below the predetermined level as noise such as dust or noise due to protruding characters. However, the blank area obtained in this way corresponds to the original cutting position of the character string pattern. Therefore, by calculating the distance L between this blank area and the end of the superimposed pattern, it is possible to determine the amount of deviation L of the temporarily set reference position. The reference position is corrected to the original reference position based on the amount of deviation L, and the character string pattern shown in FIG. 1 is cut out as shown by the broken line in the figure, whereby each character 1 is accurately cut out.

Such a series of signal processing is performed, for example, in a position detection section whose schematic configuration is shown in FIG. That is, the image signal of the character string pattern is temporarily stored in the image memory 11, and is cut out at a predetermined average array pitch as described above under the control of the cutout position calculation circuit 12 consisting of a CPU or the like. The plurality of patterns cut out in this way and outputted from the image memory 11 are guided to the projected component calculation circuit 13 to obtain their projected components, and are superimposed across the plurality of patterns. Incidentally, after the projected component calculation circuit 13 superimposes a plurality of patterns,
It may also be a method to obtain the projected component. The projected component of the superimposed pattern determined by the projected component calculation circuit 13 is then introduced into the cutout position calculation circuit 12, and the shift amount L is determined. The temporarily set reference position is corrected by this shift amount L, and the position signal is output. At the same time, the character string pattern is cut out using the corrected position as a reference.
Incidentally, the image memory 11 may store only the projected components of the character string pattern in advance, and cut out and supply the projected components to the calculation circuit 13.

Thus, according to the present device, the blank area, that is, the character gap, of a character string pattern with a known average sequence pitch is determined from the blank area of the superimposed pattern, the cutting position is determined, and the character string pattern is processed according to this. Since it is cut out, the cutout position is very accurate. Moreover, since the position is determined by determining the projected component from the superimposed pattern of the temporarily cut out patterns, the position can be detected stably and reliably even when there are characters that do not spread across the entire character frame. . Furthermore, the effects of dust and protruding characters are hardly reflected in the projected components of the superimposed pattern, resulting in S/
Good detection of N becomes possible. Therefore, with this device, there is no need to attach position detection marks to the medium as in the past, and it is less susceptible to the adverse effects of noise, so the specifications for media such as forms can be significantly relaxed, and the It also becomes possible to facilitate handling. Furthermore, if it is equipped with an automatic detection function for the average sequence pitch, it will be able to deal with various character string patterns, and its range of use will be greatly expanded. Furthermore, it is possible to simplify the next stage recognition process for each character.

Note that the present invention is not limited only to the above embodiments. For example, it is possible to adopt various methods for recognizing the average arrangement pitch, or the pitch may be determined in advance. Further, the number of patterns forming the overlapping pattern used for position detection may be determined as appropriate, and the characters are not limited to the English characters shown in the embodiments. Further, it is possible to perform position detection not only in the horizontal direction as in the embodiment, but also in the vertical direction, and it is also possible to read a two-dimensionally arranged character string pattern by using these in combination. In short, the present invention can be implemented with various modifications without departing from the gist thereof.

[Brief explanation of the drawing]

FIG. 1 is a diagram showing an example of a character string pattern with a predetermined array pitch, FIGS. 2a and 2b are diagrams showing a superimposed pattern and its projected components, and FIG. 3 is a schematic diagram of a position detecting section. 1... Character, 2... Frame, 11... Image memory, 12... Cutting position calculation circuit, 13... Projection component calculation circuit.

Claims (1)

[Claims] 1. Means for temporarily setting a cutting reference position and cutting out a character string pattern with a known average sequence pitch at the pitch, and means for superimposing a plurality of patterns cut out by this means to obtain their projected components; means for removing the influence of noise from the projected component to determine a blank area where it is determined that no characters are written; and the provisional setting based on the position of the blank area and the position of the end of the overlapping pattern. 1. A character reading device comprising: means for correcting the corrected reference position; and means for sequentially cutting out and recognizing the character string pattern at the average arrangement pitch from the corrected reference position.