JPS62169289A - Optical character reader - Google Patents

Abstract

PURPOSE:To reduce a dictionary capacity by forming a reference pattern to a partial pattern dictionary structure divided at every partial pattern, and performing a matching process by synthesizing the reference pattern with a partial pattern dictionary. CONSTITUTION:A pattern scanned at a scanning part 11, and outputted from a photoelectric converting part 12 is inputted to a feature extracting part 14 through a preceding processing part 13, and an extracted feature quantity is inputted to a matching part 15. The reference pattern in a dictionary part 17 is formed in a partial pattern dictionary structure 23 divided at every partial pattern. The matching part 15 performs the matching process by synthesizing the reference pattern for the extracted feature quantity with the partial pattern dictionary 23 using a feature synthesis instruction dictionary 22 in which the information to synthesize the partial pattern is stored.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to an optical character reading device, and particularly to an optical character reading device that is small and suitable for obtaining high reading accuracy.

[Conventional technology]

Conventionally, as a device for optically recognizing characters on a form,
Optical character readers are used. In this optical character reading device, the dictionary information prepared on the dictionary memory is created and managed for each type of character to be read (using a dedicated dictionary memory), so the number of characters that can be read increases. However, managing the dictionary memory becomes complicated,
There is a problem in that there is a lot of common dictionary information between the dedicated dictionary memories, and this becomes uneconomical in terms of effective use of the dictionary memories. In order to solve this problem, we unified the dictionary memory and specified the field read mode (whether the characters written on the form are only numbers, alphanumeric characters, etc.) when reading the form. A device is known that is capable of extracting only collated dictionary information from the unified dictionary memory at high speed (see Japanese Patent Publication No. 58-53791). In this device, dictionary information was shared in category units.

[Problem that the invention seeks to solve]

In the conventional device described above, if the dictionary configuration is as described above, in order to add more deformed patterns and standardize the dictionary, it is necessary to add an additional deformed pattern dictionary to the device, which increases the memory capacity. There was a serious problem.

The purpose of the present invention is to solve such conventional problems and to provide an optical character reading device that can significantly reduce dictionary capacity by standardizing the characteristics of partial patterns while maintaining the same level of reading accuracy as conventional devices. Our goal is to provide the following.

[Means for solving problems]

In order to solve the above problems, the present invention provides an optical character reading device having means for extracting features from an input character pattern and means for matching the feature amount extracted by the feature extracting means with a standard pattern. has a partial pattern dictionary structure divided into each partial pattern, and uses a feature synthesis instruction dictionary storing information for synthesizing the partial patterns by the matching processing means to convert the standard pattern corresponding to the extracted feature amount into the partial pattern. The feature is that it is synthesized and matched using a pattern dictionary.

[For production]

The pattern extracted by the feature extraction section is roughly divided into classes, and the class number is sent to the synthesis processing section in the dictionary synthesis section through the matching section, and the synthesis processing section uses the feature synthesis instruction dictionary to extract character feature patterns. The partial patterns are read out from the partial pattern dictionary storing the partial patterns, dictionary synthesis is performed, and matching processing is performed.

〔Example〕

Hereinafter, one embodiment of the present invention will be described in detail with reference to the drawings.

First, the present invention will be explained in detail.

The present invention was conceived in order to reduce the amount of additional mark pattern dictionaries required when reading modified patterns, and to realize a wider range of reading. One way to read deformed patterns is to expand the reading tolerance of the existing standard pattern dictionary; however,
This increases the possibility of misreading. Therefore, it is only necessary to add only the partial pattern dictionary of the modified part without changing the existing standard pattern dictionary. Therefore, the existing standard pattern dictionary is decomposed into partial pattern dictionaries, and a feature synthesis instruction dictionary is created that holds information for synthesizing the decomposed partial pattern dictionaries. As a result, partial pattern dictionaries obtained by decomposing existing standard pattern dictionaries can be made fairly common, and the number of dictionaries can be reduced significantly.

FIG. 1 is a schematic diagram of an optical character reading device showing an embodiment of the present invention.

In FIG. 1, 11 is a scanning unit that irradiates light onto the reading area of the form, 12 is a photoelectric conversion unit that receives reflected light from the form and converts it into an electrical signal, and 13 is a signal that is converted by the photoelectric conversion unit 12. A preprocessing section 14 stores the binarization result of the electrical signal in the pattern memory inside the device; 14 is the preprocessing section 1;
Characteristics such as the number of inner and outer rings and the positional relationship between line segments are extracted from the pattern for one character extracted in step 3, and the pattern is roughly classified (this is called classification, and the classification results 15 is a matching unit that performs matching processing between the feature values obtained by the feature extraction unit 14 and the standard pattern, and 16 is a matching unit that ultimately determines the character based on the matching processing result. The determination unit 17 is a dictionary synthesis unit which is the main part of the present invention.

FIG. 2 is a detailed configuration diagram of the dictionary synthesis section 17, which is a characteristic feature of the present invention.

In FIG. 2, 21 loads features from the partial pattern dictionary 23 by inputting the class number N, synthesizes the dictionaries, and
22 is a feature synthesis instruction dictionary that notifies the synthesis processing unit 21 of the storage address of the feature value corresponding to the class number N; 23 is a partial pattern obtained by decomposing the existing standard pattern dictionary; It's a dictionary. An example of this partial pattern is shown in FIG. In this embodiment, the partial pattern is a segmented line segment as shown in FIG.

The operation of the present optical character reading device will be explained below with reference to FIGS. 1 and 2.

First, the scanning section 1 irradiates light onto the reading area of a form placed at a predetermined position of the device, and the photoelectric conversion section 12 receives the reflected light from the form and converts it into an electrical signal. Here, the reflected light from the form reflects the shading information of patterns such as characters and figures on the form, and the electrical signal is a signal that corresponds to the form of the pattern. Next, the converted electrical signal is input to the preprocessing section 13.
Then, the electrical signal is binarized, the result is stored in the internal pattern memory, and noise is removed. Furthermore, in this preprocessing section 13, one
The character pattern is cut out and sent to the feature extraction section 14. Hereinafter, this pattern for one character will be simply referred to as a character pattern. The feature extraction unit 14 extracts features such as the number of inner and outer rings and the positional relationship between line segments from the input character pattern, and roughly classifies the pattern. This is called classification, and the classification result is represented by a class number N.

Next, for this class, detailed features necessary for matching are extracted based on the resulting class number N, and the matching unit 15
sent to. The matching section 15 manually inputs the class number N to the synthesis processing section 21 of the dictionary synthesis section 17 . Synthesis processing section 2
1 uses the feature synthesis instruction dictionary 22 to determine its class number N.
The storage address of the feature for the partial pattern to be synthesized is known by , and the feature is loaded from the partial pattern dictionary 23 .

Synthesize the dictionaries and obtain a signpost pattern for one character. Next, the synthesis processing section 21 sends the obtained marking pattern to the matching section 15 and performs matching with the feature value obtained by the feature extraction section 14 to obtain a degree of similarity. This process is performed for all dictionaries with the same class number N. Next, the determination unit 16 makes a final determination based on the degree of similarity and outputs the reading result.

FIG. 4 is a diagram showing a feature dictionary for partial patterns according to this embodiment. (a) is an example of a partial pattern of the modified character "2", (b) is an example of a partial pattern of another modified character "2", (c) is an example of the feature dictionary for (a), (
d) shows an example of a feature dictionary for (b).

In the case of the pattern shown in FIG. 4(a), the partial pattern S,
~S,, In the case of the pattern shown in FIG. 4(b), the partial pattern is also decomposed into ~L4, and the partial pattern features for these are μ6, ~μgm (dictionary 41), μt1~μm<
(Dictionary 42), partial patterns S1 and tl and S2 and L2 become the same partial pattern, and can be shared.
For the pattern in Fig. 4(a), μs1μs2μg, μ
S4μ6. μ66 becomes the composite feature (dictionary 41), and the fourth
μ6 for the pattern in figure (b). μ68μDμ becomes a composite feature (dictionary 43). Therefore, partial pattern dictionary μ1. It can be seen that ,μ,2 is no longer necessary, and the memory can be significantly reduced.

As shown in 2, in this embodiment, when adding a dictionary to a deformed pattern, for similar parts, only the deformed parts need to be newly added using the existing feature dictionary, reducing the man-hours required to learn a new dictionary. Moreover, since it is based on an existing dictionary, it is easy to obtain a high initial reliability of the synthesized A city pattern dictionary.

〔Effect of the invention〕

As explained above, according to the present invention, while maintaining the same reading accuracy as conventional devices, the characteristics of partial patterns are shared, so the dictionary capacity can be significantly reduced, and a compact optical character reading device can be realized. can.

[Brief explanation of drawings]

FIG. 1 is a schematic configuration diagram of an optical character reading device showing an embodiment of the present invention, FIG. 2 is a detailed configuration diagram of a characteristic dictionary synthesis section of this embodiment, and FIG. 3 is a diagram of a pattern according to this embodiment. FIG. 4, which is a diagram showing an example of decomposition, is a diagram showing a feature dictionary for partial patterns according to this embodiment. 11 scanning section, 12: photoelectric conversion section, 13 pre-processing section, 1
4: Feature extraction unit, 15 Matching unit, 16 Judgment unit,
17: Dictionary synthesis unit which is the main part of the present invention, 211 synthesis processing unit, 22. Feature synthesis instruction dictionary, 23: Partial pattern dictionary, 41. Feature dictionary, 42: Feature dictionary, 43: Commonized synthesis feature dictionary. Figure 1 Figure 2 Figure 3 Cow Figure

Claims (1)

[Claims] (1) In an optical character reading device that has means for extracting features from an input character pattern and means for matching the feature amount extracted by the feature extraction means with a standard pattern, a portion of the standard pattern divided into partial patterns is provided. Using a feature synthesis instruction dictionary which has a pattern dictionary configuration and stores information for synthesizing the partial patterns, the matching processing means synthesizes a standard pattern for the extracted feature amount with the partial pattern dictionary and performs matching processing. An optical character reading device characterized by: