JPH02202686A - Character reader - Google Patents

Abstract

PURPOSE:To increase the recognition processing speed of document contents by efficiently using plural processors. CONSTITUTION:Picture information including a character string is binarized and inputted to a picture information input means 1, and a recognition object information preparing means 2 takes out a prescribed character string area and stores recognition object information in recognition object information buffer means 31 to 3n. Character string recognizing means 41 to 4n segment character areas from character string information and extract features and collate them with standard patterns to obtain recognition results of the character string. A recognition result output means 5 outputs results of recognizing means 41 to 4n. The preparing means 2 stores picture information of the existence area of the next character string in an idle buffer even during the processing of the character string in some of recognizing means 41 to 4n when this idle buffer exists in buffer means 31 to 3n, and each of recognizing means 41 to 4n is started for the recognition processing of the next character string when terminating the processing. By this constitution, the recognition processing speed of document contents is increased.

Description

[Detailed Description of the Invention] [Summary] The present invention relates to a character reading device that has a plurality of processors and reads the contents of a document that includes image information consisting of a plurality of character string units. an image information input means for binarizing and inputting image information including character strings for the purpose of speeding up the recognition process of document contents using the image information input means; a plurality of recognition target information buffer means for respectively storing image information of the area in which the predetermined character string exists; The image characteristics of the extracted character areas are extracted and compared with standard character patterns, and the reading results of the character string are comprehensively determined from the matching results of each character area and information regarding the character string. A plurality of character string recognition means that obtain a recognition result of a character string and also perform the feature extraction and matching processing among the character string recognition processing in other character string recognition means, and from the plurality of character string recognition means. and a recognition result output means for outputting the obtained character string recognition result, and the recognition target information preparation means is configured to output the recognition result of the obtained character string even if any of the character string recognition means is in the process of recognizing the character string. , when any of the recognition target information buffer means becomes empty, the image information of the area where the next character string exists is stored in the vacant recognition target information buffer means, and the character string recognition means stores the image information of the area where the next character string exists. When any one of the processes is completed, the character string recognition means is configured to start the recognition process of the next character string.

[Industrial application field]

The present invention relates to a character reading device that has a plurality of processors and reads the contents of a document that includes image information made up of a plurality of character string units.

Character reading devices are used to optically read document images such as forms, recognize the meaning of individual characters from the information in the document image, and read the contents of the document. As the number of documents increases, there is a demand for technology that efficiently controls multiple microprocessors to achieve high-speed document reading.

[Prior art and problems to be solved by the invention]

The general structure and operation of a conventional character reading device using a plurality of microprocessors is as follows, for example.

FIG. 7 shows the hardware configuration of the main parts of a conventional character reading device.

In FIG. 7, 11 is an image memory, 12 is a control 'B processor, 13 is a recognition target information buffer, 141, 14
□ is a recognition processor, and 15 is a form format memory.

The image memory 11 stores, for example, binarized image information of a document input by an optical image reading device (scanner) or the like.

For example, when the document to be read is a form, the form format memory 15 holds information about the form, such as what information is written in which position and what kind of character string is written on the form. ing.

The control processor 12 stores the form format memory 1
Based on the information held in the image memory 11 , image information of an area corresponding to a predetermined character string is extracted from the image information stored in the image memory 11 , and is set on the recognition target buffer 13 . , 14□ starts one of the character string recognition operations.

The recognition processor that received the above activation, for example, 14°,
First, individual character areas are cut out from the image information of the M area corresponding to a predetermined character string set on the recognition target buffer 13. Then, the features of the image of multiple character regions cut out from the region of the character string are extracted, and compared with standard character patterns in a dictionary (not shown), and the results of matching all the characters are compared to the predetermined character. The 1ll17 result of the character string is determined by making a comprehensive judgment along with the information regarding the string. here,
The other recognition processor 14z performs the matching process of the images of the plurality of character areas together with the recognition processor 14°.

The obtained character string recognition result is then output to other processing stages via the control processor 12.

In the conventional character reading device shown in FIG. It was not possible to proceed to the next operation until the recognition result of the character string for the activation was obtained.

Therefore, the addition of each character string v2 recognition means varies depending on the layout of the form and the state of writing of characters, and there is a problem that there is a lot of idle time due to waiting for processing, and the overall document reading speed decreases.

The present invention has been made in view of the above problems, and an object of the present invention is to provide a character reading device that efficiently uses a plurality of processors to speed up document content recognition processing.

[Means to solve the problem]

FIG. 1 is a basic configuration diagram of the present invention. In this figure, ■
2 is image information input means, 2 is recognition target information preparation means, 32
32, . . . 3n are recognition information buffer means, 4□4n,
... 4n is a character string recognition means, and 5 is a recognition result output means.

The image information input means 1 binarizes and inputs image information including character strings.

The recognition target information preparation means 2 extracts an area in which a predetermined character string exists from the input image information.

The recognition information buffer means 3n32, . . . 3o each store image information of the area where the predetermined character string exists.

Character string recognition means 41142! ...4n cuts out individual character areas from the image information of the stored character strings, extracts the features of the image of the cut out character areas, matches them with a standard character pattern, and calculates the matching results of each character area and The recognition result of the character string is obtained by comprehensively determining the reading result of the character string from the information regarding the character string, and the above-mentioned characteristics of the character string recognition processing in the other character string recognition means 4n42...4o Extraction and verification processing will be carried out separately.

The recognition result output means 5 includes the plurality of character string recognition means 4I.
, 42 . . . , output processing of the recognition results of the character strings obtained from 41 is performed.

Furthermore, the recognition target information preparation means 2 can recognize any one of the recognition target information buffer means 3I+ 3t+ ... 3R even if any of the character string recognition means 41+ 4t+ ... 4n is in the process of recognizing a character string. When an empty space occurs, the image information of the area where the next character string exists is stored in the recognition target information buffer means of the empty space, and the processing of any one of the character string recognition means 44142...4n is performed. When this is completed, the character string recognition means is activated to recognize the next character string.

[For production]

As an example, FIG. 2 shows the comparison with the conventional character reading device described above when performing the recognition processing of character strings ■, ■, and ■ in order when there are two character string recognition means. , which shows how the processing time is shortened in the character reading device of the present invention.

In FIG. 2, "preprocessing" of "control P" corresponds to the operation of the recognition target information preparation means 2 of the present invention (FIG. 1),
"Post-processing" in "Control I8 PJ" indicates the operation of the recognition result output means 5 of the present invention. These also correspond to the operations of the control processor 12 of the conventional configuration (FIG. 7) described above.

Furthermore, in FIG. 2, "recognition P#l and #2j correspond to the character string recognition means 41 and 42 of the present invention (FIG. 1), and also correspond to the character string recognition means 41 and 42 of the above-mentioned conventional configuration (FIG. 7). It corresponds to the recognition processor 14□14□.

In the conventional configuration shown in FIG. 7, there is only one recognition target information buffer (13 in FIG. 7) that stores the image information of the area corresponding to the character string. While recognition processing is being performed by recognition P#1 and #2, processing for storing image information of the area corresponding to the next character string (ie, preprocessing) cannot be started.

However, according to the present invention, a plurality of recognition target information buffer means are provided, and if any of the recognition target information buffer means has a free space, the recognition target information preparation means 2 (
1), image information of an area corresponding to the next character string can be stored in the empty recognition target information buffer means even during recognition processing of the previous character string. The character string recognition means can also be activated to recognize the next character string as soon as the feature extraction and matching processes are completed. That is, as shown in Figure 2,
Even if any of the character string recognition means (recognition P) is in the process of recognizing the previous character string, the next character string can be preprocessed.

〔Example〕

FIG. 3 schematically shows the hardware configuration of the main parts of the embodiment of the character reading device of the present invention.

FIG. 3 corresponds to FIG. 7 showing the above-mentioned conventional configuration, and in FIG. retains the functionality of the corresponding configuration.

However, in the configuration of FIG. 3, according to the present invention, a plurality (n) of recognition target information buffers are provided, and furthermore, the control processor 12 realizes the function of the recognition target information preparation means 2 of FIG. As shown in FIG. 1, if there is a free space in any of the recognition target information buffers 132132, image information of the area corresponding to the next character string is stored in the free recognition target information buffer even during recognition processing of the previous character string. do. The recognition processor corresponding to the character string recognition means shown in FIG. 1 can also be activated to recognize the next character string as soon as the feature extraction and collation processes are completed. That is, even if any of the recognition processors is in the process of recognizing the previous character string, the next character string can be preprocessed.

Furthermore, the control processor 12 in FIG. 3, when a recognition result for one character string is obtained in any recognition processor, regardless of the operation of other recognition processors,
A process (post-processing) is performed to read out the recognition result and transfer it to another area.

The timing of such processing is as shown in FIG.

FIGS. 4 and 5 show an example of the control flow in each of the control processor 12 and recognition processors l'L and 14z configured as shown in FIG. 3.

First, in the control processor of FIG.
1, if any of the recognition target buffers is usable (empty), and furthermore, in step 402, if image information of an unprocessed character string exists on the image memory 11, steps 403 to 403 described below are performed. 405 preprocessing is performed.

First, in step 403, information regarding the above character string on the form format memory 15 is read, in step 404, an area of the character string (field to be recognized) is cut out from the image memory 11 according to this information, and in step 405, the area of the character string (field to be recognized) is extracted from the image memory 11. , the data is converted into a data format for processing by the recognition processor and set in the empty recognition target information buffer.

Then, in step 406, if any recognition processor is available, in step 407, character string recognition processing starting from character extraction processing is activated for the vacant recognition processor.

Furthermore, in step 408, after the above startup, the recognition processor waits for the extraction of one character to be completed, and in step 409, another recognition processor performs feature extraction and matching processing (hereinafter referred to as matching) for the one character. (abbreviated as processing)
If it can be used to share the processing, step 41
At step 0, the above-mentioned one character is transferred to another recognition processor, and at step 411, matching processing in the recognition processor is started.

In the above procedure, if in step 401 none of the recognition target buffers is usable (empty) or in step 4
If there is no unprocessed image information on the image memory 11 in step 02, the process advances to step 412.

Further, if any recognition processor is disabled in step 406, or if another recognition processor is disabled in step 409, the process returns to step 401.

In step 412, if any recognition processor has obtained a character string recognition result, the process proceeds to steps 413 and 414 for post-processing. That is, in step 413, the recognition result is read out and transferred to another area, and in step 4
At step 14, various control parameters related to the character string are updated. As a result, the recognition target information buffer that was being used for the character string becomes empty, and the recognition processor that was being used becomes available for use.

In step 412, if none of the recognition processors has obtained a character string recognition result, in step 415, it is determined whether the matching process is being divided between the two recognition processors, and the process is divided. When it is determined that processing is in progress,
If a matching result has been obtained in the recognition processor on the sharing side (the other recognition processor) at step 416, the matching result is transferred to the character string recognition processor (character string recognition processor) in step 417. In step 418, if there are any characters for which matching processing has not yet started (cut out), the process returns to step 409.

If it is determined in step 415 that the matching process is not being shared between two recognition processors, or if the matching process result for the sharing side has not yet been obtained in step 416, step 419 is performed. , it is determined whether there is a recognition target information buffer for storing unprocessed data, and if there is, the process proceeds to step 406; if not, the process proceeds to step 401.

In the recognition processor that operates under the control of the control processor 12, first, in step 501 of FIG. 5, if the control processor starts the character string recognition process, in step 502, ( The image information of the character string area stored in the recognition target information buffer (specified by the control processor at the time of startup) is read out, and each character area is cut out from the image information. And step 50
In step 3, the features of the cut out characters are extracted, and in step 50
In step 4, comparison with the standard pattern is performed. Then, in step 505, the matching results are accumulated.

Next, in step 506, under the control of the control processor, if there is a result of the matching process shared by another recognition processor, this is also accumulated in step 507.

Thereafter, the processes of steps 503 to 507 are repeated until the collation process of all the characters cut out in step 502 is completed, and when the process is completed, the meanings of these character strings are stored in a knowledge dictionary (not shown) or the like in step 509. The character string is comprehensively determined while referring to the character string, and the determination result for the character string is obtained, and in step 510, this is transferred to the control processor.

On the other hand, when the other recognition processor is activated by the control processor to share and collate the characters cut out as described above, the control processor receives the transfer of the cut out characters at the Ste knob 512, At step 513, the features are extracted, and at step 514, a comparison with a standard pattern is performed.Furthermore, at step knob 515, the matching result is transferred to the other recognition processor.

〔Effect of the invention〕

According to the character reading device of the present invention, a plurality of processors can be used efficiently to speed up the recognition process of document contents.

[Brief explanation of the drawing]

FIG. 1 is a basic configuration diagram of the present invention, FIG. 2 is an explanatory diagram of the operation of the present invention, FIG. 3 is a schematic hardware configuration diagram of an embodiment of the present invention, and FIG. 4 is a control flow diagram of the control processor. FIG. 5 is a diagram showing an example of the control flow of the recognition processor; FIG. 6 is a timing diagram of the configuration shown in FIG. 3; and FIG. 7 is a schematic hardware configuration of a conventional character reading device. It is a diagram. [Description of symbols] ■...Image information input means, 2...Recognition target information preparation means, 31, 32,...31...Recognition information buffer means, 4n42...4i...Characters Column recognition means, 5.
... Recognition result output means, 11... Image memory, 12.
... Control processor, 13n131.13n... Recognition target information buffer, 14,, 14□... Recognition processor, 15... Form format memory.

Claims (1)

[Claims] 1. Image information input means (1) that binarizes and inputs image information including a character string, and recognition target information that extracts an area where a predetermined character string exists from the input image information. a preparation means (2); and a plurality of recognition target information buffer means (3_1, 3) each storing image information of an area where the predetermined character string exists.
_2,...3_n), cut out individual character areas from the image information of the stored character strings, extract the features of the image of the cut out character areas, and compare them with standard character patterns to determine the characteristics of each character area. A recognition result of the character string is obtained by comprehensively determining the reading result of the character string from the matching result and information regarding the character string, and other character string recognition means (4_1, 4_2...4_n)
A plurality of character string recognition means (4_1,
4_2...4_n), and the plurality of character string recognition means (4_1, 4_2...4_
and a recognition result output means (5) for outputting the recognition result of the character string obtained from the character string recognition means (4_1, 4_
2...4_n), the recognition target information buffer means (3_1,
3_2,...3_n), the image information of the area where the next character string exists is stored in the vacant recognition target information buffer means, and 4_1, 4_2, .