JPS6073793A - Optical character reader - Google Patents

Abstract

PURPOSE:To enable the titled reader to recognize characters by the same pre- treatment means and the same dictionary even when the inversion characters are mixed by inverting the inversion character to return it to the pre-treatment part when the inversion character is detected. CONSTITUTION:The pre-feature monitor part 4 monitoring a cut-out image data while cutting out the image data from the operation buffer 6, decides the feature of the easy category that there is no concaved or a cut part at the external border, thereby detecting the inversion image data. When the data is decided to be the inversion image data, its white and black are inverted at the inversion treatment part 9 and the data is returned to the pre-treatment part 5 again, and is binary encoded. Therefore, the dust is removed and smoothing is executed because the data is already binary-coded, then the data is returned to the buffer 6 again. For this reason, the monitor part 4 will work only as an ordinary cut- out control, recognition treatment and editing treatment are executed at the ordinary route, and as a result, the data is outputted to the data memory 11.

Description

DETAILED DESCRIPTION OF THE INVENTION Field: The present invention relates to an optical character reading device, and more particularly to a device having a reflex processing loop for recognizing inverted characters as well as normal characters.

<Prior art and background> So-called optical character reading devices (commonly referred to as OCR devices) are used for machine recognition of characters and figures, but the objects they usually recognize are printed or handwritten on media whose substantial parts are black. It is a character figure that has been However, when considered as a data input mediating means for an information processing system, the data expression format of the form that is the medium is a special expression 2 due to headings and strong expressions, such as white letters with black and white reversed, or double strokes that are slightly shifted and overlapped. There is a demand for using characters, etc.

When the means for writing data on the medium is a so-called matrix printer, the normal pattern and the white pattern can be created using the same pattern memory contents of the pattern generator, and are therefore relatively frequently used.

On the other hand, when considering this as a recognition target from the standpoint of an OCR device, the appearance rate is not very large, but if a proper dictionary is prepared, at least twice as many dictionaries are required, and the collation time for recognition becomes longer. Also, when the read image information is inverted (white out), voids (white as noise) tend to appear on solid black areas, but the preprocessing to fill in these voids is different from the preprocessing for normal pattern information. It ends up being the opposite.

<Purpose and Features> In view of the above, the object of the present invention is to perform inversion processing on the pattern data caught at the time when an inverted character is caught and return it to the preprocessing section, thereby preprocessing it by the same preprocessing means and using the same dictionary. The purpose of the present invention is to recognize inverted characters, and the features of the present invention include a preprocessing unit for recognizing and processing the image information read for the above purpose, a cutting unit, a recognition processing unit, and a main controller that processes these data and controls the means. The optical character reading device has a means for detecting that the information is inverted when cutting out the information, and a working unit that inverts the cut out information, and if it is inverted information, the cut out information is inverted and read again. This is to control the return to the processing section.

<Embodiment> FIG. 1 is an explanatory diagram of an embodiment of the present invention. In the figure, 1 is a form, 2 is a light source, 3 is an optical selector unit that reads the information image (original read data) from the form, 4 is a pre-feature monitoring unit that normally performs cutting and detects inverted information, and 5 is a unit that detects inverted information. The information image is temporarily stored as image data, the image data is binarized, and further preprocessing for recognition processing such as so-called dust removal and smoothing is performed, and the processed image data is stored in the work buffer 6.
7 is a dictionary for recognition, 8 is a recognition working unit, and 9 is an inversion processing unit that converts the image data in the work buffer 6 into inverted data.

10 is a control memory for storing format information of form 1 as control information; 11 is a data memory for storing recognized and edited processed recognition data; 12 is for editing recognized data and controlling these means. This is the main control unit for performing the following: The thick line indicates the flow of data, and the thin line indicates the correspondence of control information.

2, 3, and 4 are auxiliary diagrams of FIG. 1, and FIG. 2 is for explaining the working configuration of the preprocessing section 5, and FIG. 3 is for explaining the processing flow. It is something. Further, FIG. 4 is a flowchart of an example of the previous feature determination work at the time of extraction in the previous feature monitoring unit 4.

As is clear from the figure, normal image data is processed by the preprocessing unit 5.
The characters are processed and stored in the work buffer 6, and are extracted one character at a time by the monitoring section 4, which also serves as extraction control.The recognition section 8 refers to the contents of the dictionary 7 to perform character recognition, and the recognized results are sent to the main control section. 12 and stored in the data memory 11. Note that the format information in the control memory 10 is used to control the extraction of data before recognition and to edit the data after recognition into an output format. The configuration and work flow up to this point are the same as the conventional one. However, this embodiment has a pre-feature monitoring section 4 which monitors the cut-out image data while cutting out the image data from the work buffer 6. It detects that the image data is not normal image data but is inverted image data by making feature judgments in very simple categories such as the ratio of large black is larger than a predetermined value, there are no dents or cuts in external links, etc. do. Then, what is determined to be inverted image data by this detection is inverted in black and white in the inversion processing unit 9, and returned to the preprocessing unit 5 again.Since the binarization has already been performed, it is subjected to dust removal and smoothing, and then transferred to the work buffer 6 again. return. By the way, since the image data has already been inverted at this time, the front feature monitoring section 4 operates only as normal extraction control, and the recognition processing and editing processing is performed in the normal route, and the result is output to the data memory 11. .

With this configuration, it is possible to perform the same category format confirmation judgment using the same preprocessing unit and the same dictionary as before, and to recognize images without worrying about the expression format even when normal format image data and inverted format image data are mixed. I can do it.

Of course, the processing time required to go through the return loop will be extra, even for simple judgments, but as a normal system operation mode, the frequency of appearance of characters for expressions for emphatic expressions is naturally lower than that for normal expressions. Since the frequency of character appearance is low compared to the character appearance frequency, even if such reflex route processing is performed, the overall system output capacity does not deteriorate much, and the hardware configuration is simple and the implementation cost is low. The benefits that can be realized should receive more attention.

Note that these data processing and procedure control functions can be replaced with internal programs of the microprocessor.

As a variation of this embodiment, it is also possible to have the pre-monitoring section handle the throughput processing. <Effects> As explained above, according to the present invention, the capacity of the dictionary does not particularly increase, and it is not necessary to use hardware or microprograms. This has the characteristic effect of providing a character recognition device that can commonly recognize normal characters and inverted characters without increasing the number of characters too much.

[Brief explanation of drawings]

FIG. 1 is an explanatory diagram of one embodiment of the present invention, and FIGS. 2, 3, and 4 are supplementary explanatory diagrams of FIG. 1. In the figure, 4 is a pre-feature monitoring section, 5 is a pre-processing section, 6 is a work buffer, 7 is a dictionary, 8 is a recognition work section, 9 is an inversion processing section, 10
is a control memory, 11 is a data memory, 12 is a main control unit,
5a is a binary conversion section, 5b is a dust removal section, and 5c is a smoothing section. Figure 3 Figure 4

Claims (1)

[Claims] In an optical character reading device having a pre-processing section for recognition processing of read image information, a cutting section, a recognition processing section, and a main control section for processing these data and controlling procedures, the inverted information is used for the above-mentioned cutting. An optical character reading device comprising a means for detecting this and a working part for inverting the cut out information, and controlling the cut out information to be inverted and returned to the pre-processing part again if the cut out information is inverted information. .