JPH08293003A - Character string extraction method, character string extraction apparatus using the method, character string recognition processing apparatus and character string recognition processing system using the apparatus - Google Patents

Abstract

(57) [Summary] [Purpose] To accurately extract the character string to be recognized. [Structure] When a model of a character string to be extracted is set by the image data setting unit 1, the image cutout unit 2 cuts out an image of a predetermined region from the image data of the target object and supplies it to the character extraction unit 3. . When the character extracting unit 3 extracts all the characters existing in the given image and their characteristic amounts,
The classification processing unit 4 divides each continuous character into groups using the characteristic amount of each character and extracts the characteristic amount of each group. The comparison processing unit 5 compares the feature amount data of each group with the feature amount data T of the model image, and determines the group having the feature amount data closest to the model as an extraction target. The feature amount data of the determined group is given to the output processing unit 6, where the data relating to the character string is extracted and output.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method and apparatus for extracting a specific character string from an image containing a plurality of characters, and the present invention further relates to an object having a plurality of character strings. The present invention relates to an apparatus and a system for extracting each character string from an image and performing recognition processing.

[0002]

2. Description of the Related Art In recent years, an apparatus has been developed which performs a recognition process by extracting a specific character string from an image obtained by picking up an image of an object on which various character strings are written. In this type of device,
After setting a predetermined processing area containing the character string to be processed on the image of the object, the characters contained in the area are sequentially extracted using a method such as pattern matching, and the character string recognition processing is performed from the extraction result. It is carried out.

[0003]

However, in the above method, since all the characters included in the area are the recognition result, if the area contains characters other than the recognition target, the characters are also extracted. , The recognition result is incorrect. For this reason, it is necessary to strictly determine the setting position and size of the processing area, but if the positioning accuracy of the object is poor or if there are many characters other than the recognition target in the image, the target character It is difficult to set the area so that the columns are accurately extracted, and the recognition accuracy is significantly reduced.

By the way, as a method for extracting a character string written along a diagonally extending line segment, first, the inclination angle of the line segment is obtained, and the characters in the vicinity of the line segment are sequentially extracted according to the inclination angle. Has been proposed (JP-A-62-281).
086 publication). By applying this method, even if the character string is inclined, the inclination angle can be obtained to extract the character. However, in this case as well, if the image includes characters other than the recognition target, it is difficult to obtain the tilt angle in character string units, and therefore the recognition accuracy cannot be reduced.

The present invention has been made in view of the above problems, and a plurality of characters included in an image are grouped into groups for each character string having continuity, and the characteristic amount of each group is set in advance. An object of the present invention is to provide a character string extracting method and an apparatus for accurately extracting only a character string to be recognized by comparing it with a model of a character string feature amount.

Further, according to the present invention, the above processing is executed for an object having a plurality of character strings such as a business card, and the extracted character strings are recognized to accurately grasp the information contained in each character string. An object of the present invention is to provide a character string recognition device and a system thereof.

[0007]

[Means for Solving the Problems]

<Technical idea devised to solve the problem> The above-described object of stably extracting a character string to be processed from an image including a plurality of character strings even if the image is rotated or translated. The technical ideas devised to achieve this are as follows. That is, a plurality of characters included in the image are grouped into character strings having continuity, and the feature amount of the character strings in each group is compared with a model of the feature amount of the character string registered in advance, and the extraction target is extracted. The character strings to be processed are extracted by discriminating the groups.

Based on this technical idea, the method described in claims 1, 2 and 3 was used as the character string extraction method, and the means described in claims 4, 5 and 6 was used as the character string extraction device. Furthermore, the means described in claim 7 is used as a character string recognition processing device, and a character string recognition processing system is used.
The procedure described in 1. was used.

<Definition of Special Terms Described in the Claims> The meanings of the special terms described in the claims are clarified as follows, and corresponding portions in the embodiments will be described.

"Character string having continuity" It is a character string that is substantially located on a predetermined straight line or curved line. In the embodiment, character strings arranged on a straight line at substantially equal intervals are set as a character string having continuity. However, the character string having continuity is not limited to a straight line, and may be a character string arranged on a curved line such as an arc or a parabola.

[0011]

The plurality of characters contained in the image are grouped into groups of character strings having continuity. The feature amount of each group is compared with the model of the feature amount of the character string to be extracted, and it is determined that the group having the feature amount closest to this model corresponds to the model. By outputting the information of the characters included in this group, the target character string is accurately extracted.

According to the second and fifth aspects of the present invention, as long as the relationship between the two characters is such that the difference in the arrangement direction of the characters on the image falls within a predetermined error range,
These letters fall into the same group.

In the inventions of claims 3 and 6, since data relating to the size of the character is used as the feature amount to be compared between each group and the model of the character string, the character string to be recognized is misaligned. You can easily and accurately extract character strings even when rotating or rotating.

According to the invention of claim 7, after extracting all the characters included in a predetermined range from the image obtained by imaging the object having a plurality of specific character strings, these characters are consecutive. Are grouped into groups for each character string having. By comparing the feature amount of each group with the model of the feature amount of each character string, the group corresponding to each model is determined, and the character information of these groups is recognized and processed as the character string to be processed.

According to the invention of claim 8, a plurality of characters are extracted from images obtained by sequentially capturing a plurality of objects having a plurality of specific character strings, and are grouped in the same manner as described above. After that, a group corresponding to the model of the characteristic amount of each character string is determined. After the character information of these groups is individually recognized, the recognition result is recorded for each object and read as necessary.

[0016]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS FIG. 1 shows the structure of a character string extracting device according to an embodiment of the present invention. This character string extracting device takes in an image of an object in which a character string is written at a predetermined position such as paper, and extracts the character string from the image. The model image data setting unit 1, the image cutting unit 2, Character extraction unit 3,
It includes a classification processing unit 4, a comparison processing unit 5, an output processing unit 6 and the like as its configuration.

The model image data setting unit 1 is for setting a model image of a character string to be extracted prior to extraction processing, and further extracts feature amount model data T described later from this model image, Store in internal memory.

When the model image is set, the image cutting unit 2 then shifts to the image cutting process. In this process, image data of an object is captured from an imaging device (not shown) or another image processing device, and an image of a predetermined area including a character string to be recognized is cut out from this image. It is output to the extraction unit 3. The cut-out position of this image is obtained by a method such as being set by an operator in advance or being extracted from the data when setting the model image.

The character extraction unit 3 has a memory for storing various character font data, and extracts all the characters existing in a given image using a pattern matching method. This font data has data for each character whose rotation angle is 0 degree (hereinafter referred to as "reference font data") and a plurality of data rotated by a predetermined angle (hereinafter referred to as "rotated font data"). By matching the reference font data and the rotation font data of each character with respect to all the characters existing in the image, the character type and rotation angle of each character are extracted.
Further, the character extraction unit 3 extracts data relating to the position and size of each character for each extracted character, and then summarizes the feature amount of each character in the form of a vector (hereinafter referred to as “feature amount data”). Output) to the classification processing unit 4.

FIG. 2 shows an example of the above feature amount extraction.
In the illustrated example, as a result of performing the above-described pattern matching on the character L _i in the image cut out by the image cutout unit 2, the character type c _i (“1” in the illustrated example) and the tilt angle a _i are extracted. There is. The character extraction section 3, extracted slope a _i predetermined rectangular area r _i including characters in accordance with the
, And then the x, y of the lower left vertex p _i of this region r _i
The coordinates (x _i , y _i ) are extracted as the position data of the character L _i . Furthermore the lateral sides of the region r _i as the width of the character length w _i
But the length h _i of the vertical sides of the region r _i is the height of the characters,
Each is extracted.

[0021] in the form of this at the feature data M _i of the character L _i vectors, it is as follows (1).

[0022]

[Equation 1]

In the equation (1), g _i is a parameter for grouping described later (hereinafter referred to as "group data"), and "0" is set in the initial state.

Returning to FIG. 1, the classification processing unit 4 checks the continuity of each character using the feature amount data of each character provided from the character extraction unit 3 according to the procedure described later, and for each successive character. Divide into groups. Further, the classification processing unit 4
The feature amount of the group is obtained using the feature amount data of the characters belonging to each group, and the feature amount data obtained by collecting the feature amounts in the form of a vector is generated and output to the comparison processing unit 5.

The comparison processing unit 5 reads the feature amount model data T from the model image data setting unit 1 and sequentially compares the given feature amount data for each group with the feature amount model data T. As a result, the group having the feature amount data closest to the model image is determined to be the character string to be extracted, and the feature amount data of the group is output to the output processing unit.

The output processing unit 6 extracts an element related to the character type from the feature amount of the given group and outputs this as an extraction result to the outside.

3 (1) and 3 (2) show the principle of the grouping process by the classification processing unit 4. In this example,
It is premised that the characters L _k and L _S are extracted and the characteristic amount data M _k and M _{S of} each character are extracted by the same procedure as the above-mentioned character L _i . In the figure, x _k , y
_k is the position data of the character L _k (coordinates of the vertices p _k region r _k containing the character _{L k), x S, y} S vertex region r _S including the position data (character L _S letter L _S p The coordinates of _S ) are shown.

As described above, since the inclination of the first character L _i is a _i , the linear equation in which this character L _i is located is expressed by the following equation (2).

[0029]

[Equation 2]

If the second character L _k is a character following this character L _i , the character L _k should exist on the straight line represented by the above equation (2). Therefore, the necessary and sufficient condition for the character L _k to continue to the character L _i is expressed by the following equation (3). Note that ε in the formula is a preset error.

[0031]

(Equation 3)

Next, if the third character L _S is a character continuous with the character L _k , it is considered that the successive characters are arranged at substantially equal intervals, so the position of the character L _S is (x ₂ +
(X ₂ -x _1), it should be near the point represented by _{y 2 + (y 2 -x 1} )). Therefore, the letter L _S is the letter L
Necessary and sufficient conditions for continuation of _i and the character L _k are expressed by the following equations (4) and (5).

[0033]

[Equation 4]

[0034]

(Equation 5)

The characters L _k and L _S are _{defined in the} above (3)-
When the condition of expression (5) is satisfied, three characters L _i ,
L _k and L _S are determined to be continuous and are classified into the same group. The classification processing unit 4 further, for this group, the feature amount data M _i , M _k , M of each character described above.
After reading _S , the average value of the width and height of the characters is calculated, and the distances d ₁ and d ₂ between the characters (shown in FIG. 3 (2)) are calculated to calculate the average value. Feature value data in the form of a vector with values and the character type of each character is created.

[0036] For now j-th group G _j, the average distance md _j between each character, the average character width mw _j, the average character height mh _j is assumed to be calculated, the feature quantity of the group G _j The data CG _j is represented by the vector shown in the following equation (6).

[0037]

(Equation 6)

Α _j is data indicating a character string in which the character types in the group G _j are arranged in the order of extraction (“123” in FIG. 3), and n _j is data indicating the number of characters belonging to the group. (“3” in the case of FIG. 3).

FIG. 4 shows a control procedure of the character string extracting device. In the following description, N represents the number of characters extracted by the character extraction unit 2, and NN represents the total number of groups set by the classification processing unit 4.

First, step 1 (indicated by "ST1" in the figure)
Then, when the model image data setting unit 1 sets the image of the character string that is the model of the character string to be extracted, the image cutout unit 2 takes in the image of the target object and extracts the image data of a predetermined area from it. It is taken out and output to the character extraction unit 3 (step 2).

Next, the character extraction unit 3 sequentially scans the given image data and performs pattern matching processing to determine the character type, inclination, position data, character width, and character height of each character included in the image. Each feature amount is extracted, and an initial value "0" is set as the group data (steps 3 and 4). These feature quantities are put together in the form of the vector of the above-mentioned expression (1), and are output in the ascending order of the x coordinate of the position data (in the ascending order of the y coordinate if the x coordinate is the same value). It is stored in the internal memory inside.

[0042] Figure 5 shows a storage state of the feature amount data of each character, the feature amount data M ₁ ~M _N of the extracted N characters L ₁ ~L _N, each vector element is set Are sequentially stored.

Returning to FIG. 4, the classification processing unit 4 sets the counter i for specifying the character to be classified in step 5 and the group data g _i of the character specified in step 6. The counter j is set to the initial value "1", and the processes of steps 7 to 19 are repeated to extract consecutive characters, and each character is grouped based on the extraction result.

First, the classification processing unit 4 determines in step 7 whether the counter i has exceeded the total number N of extracted characters. Since the counter i is "1" at the present stage, the determination in step 7 is "NO" and the process proceeds to step 8
The feature amount data M ₁ (shown in FIG. 5) of the character L ₁ is read and it is checked whether or not the group data g ₁ is “0”.

As described above, since the group data g ₁ is set to "0" at the initial stage, the determination in step 8 is "YES", and the classification processing section 4 determines the value of the group data g ₁ The value of the counter j is set as. Since the counter j is “1” at this stage, the first character L ₁ is classified into the first group in this case.

Next, the classification processing section 4 determines x in the equation (3).
_The currently read position data (x ₁ , y _{1 in} this case) is substituted into _i and y _i, and the feature amount data M _{2 to MN} of other characters are sequentially read, and the position data is stored in the above (3). (3) is executed by substituting it into x _k and y _k of (1) and it is checked whether the obtained value is less than or equal to the error ε (step 10). As a result, if there is a character satisfying the condition, the character is specified as a character L _k that is continuous with the character L _i, and j is set as the group data g _k of the character L _k (steps 11 and 12). When a plurality of characters satisfying the expression (1) are found, the character closest to the character L _i is specified as the character L _k .

Further, the classification processing unit 4 substitutes the position data of the characters L _i and L _{k into} the expressions (4) and (5), and the position data of the characters other than these two characters are sequentially x in the expression (4). _S, (5) running y _S to assign to two equations of equation to check for characters these results are both less error epsilon. When a character satisfying this condition is found, it is specified as a character L _S that is continuous with the character L _k , and this character L
To set the value of the counter j as a group data g _S in _S (step 14, 15). When this processing ends, the classification processing unit 4 sets the character L _S to L in order to check whether there are more consecutive characters in the next step 16.
_It is replaced with _k, and the process returns to step 13 again to execute the same processing.

When the determination in step 14 is "NO", that is, when the character L _S consecutive to the character L _k is not found, the classification processing unit 4 determines in step 17 that the value of the counter i is N or less. After confirming that, the counter j is incremented in the next step 18.

After the increment processing in step 18,
Alternatively, when the determination in step 11 or 17 is "NO", the counter i is incremented in step 19 and the processing from step 7 onward is executed again for the new character L _i . In this case, if this character L _i has already been processed as being continuous with other characters, the value of “1” or more is set in the group data g _i , so the determination in step 8 is “NO”. , Steps 9 to 18 are skipped.

When the above process is performed for all characters, the determination in step 7 is "YES" and step 2
Then, the classification processing unit 4 sets the final value of the counter j to the total number of groups NN. Furthermore, the classification processing unit 4
Collects the feature amount data of each character in the memory for each group having the same group data to generate the feature amount data of the group shown in the equation (6), and sequentially outputs the feature amount data to the comparison processing unit 5.

The feature data CG ₁ ~CG _NN of each group G ₁ ~G _NN output, as shown in FIG. 6, are sequentially stored in the internal memory of the comparison processing unit 5. The comparison processing unit 5
In addition, stores in other areas of the memory reads the feature amount model data T from the model image data setting unit 1, of the elements of the feature amount data CG _j for each group G _j, the number of characters n _j, The average distance md _j between characters, the average character width mw _j , and the average character height mh _j are given by the following (7) to (10).
The feature amount model data T is sequentially compared based on the condition of the formula (step 22). Note In the following formulas, n _t, md _t, m
w _t and mh _t are elements for the feature amount model data T and indicate the number of characters, the average distance between characters, the average character width, and the average character height, respectively.

[0052]

(Equation 7)

[0053]

(Equation 8)

[0054]

[Equation 9]

[0055]

[Equation 10]

In the above equation, εw and εh are allowable errors in the width and height of the character, respectively, and the comparison processing unit 5 determines that the feature amount of the group that satisfies all of the above equations is the smallest. A group close to the model data is determined as an extraction target group, and the characteristic amount data is output to the output processing unit 6.

The output processing unit 6 extracts the element α related to the character string from the feature amount data of the given group and outputs it as the extraction result (step 23).

According to the above character string extraction method, a character string written in a predetermined place of an object, such as an examination number on a test answer sheet or a serial number attached to various products, can be easily and accurately obtained. It is possible to extract.

Further, the character string to be extracted is not limited to one string, and a plurality of character strings can be extracted. In this case, individual model image data is set for each character string in advance, characters are extracted from the image of the target object and grouped in the same manner as described above, and then the feature amount data of each group are respectively set. Compared to the feature model data of
It suffices to extract the groups corresponding to each model and extract the elements related to the character strings of these extracted groups. According to this method, it is possible to construct a database by reading each information from an object having a plurality of data such as a business card and a customer card.

FIG. 7 shows a schematic structure of a database system for managing the information of business cards. The image scanner 7, the control device 8, the magnetic disk device 9, and the monitor 1 are shown.
0, a keyboard 11 and the like are included as components.

The control device 8 includes a CPU 12 and a ROM 1
3, the image input board 15 in the control unit composed of RAM14
(Including A / D conversion unit 16 and image memory 17), disk interface 18, I / O interface 19
Are connected via a bus. Note that the CPU 12 has all the functions other than the memory among the functions of the respective units in FIG. 1, and the RAM 14 stores the characteristic amount data, the calculation result, and the like for the characters and groups described above. Further, the ROM 13 stores a series of control programs, and also stores standard font data and rotation font data for each character such as kana, kanji, numbers, and English characters specified for the above-described character extraction processing.

The image data of the business card generated by the image scanner 7 is input to the A / D converter 16 of the image input board 15 and digitally converted, and then stored in the image memory 17. The CPU 12 takes out the image data in the predetermined area from the image memory 17, extracts all the characters contained therein, and then, based on the above-mentioned conditions,
Each character is divided into groups, and the characteristic amount data of each group is stored in the RAM 14.

The magnetic disk device 9 stores a model data file 20 for storing the feature amount of the model image of each character string to be recognized, and the result of processing each character string to be recognized. The database file 21 is set. FIG. 8 is an example of the data structure of the model data file 20, and is a feature quantity model such as a character type, a character interval, a character width, and a character height for each data type described in a business card such as a name and an address. The data is set.

In the above feature amount model data, the character type indicates the type of characters (Kanji, Kana, numbers, English characters, etc.) that make up the character string. The character spacing, the character width, and the character height are the same data as in the first embodiment, and the maximum value (DN, DA, ...) And minimum value (dn, (da, ...) Is stored.

The CPU 12 sequentially reads the feature amount data of each group from the RAM 14, compares the feature amount data with each model data of the model data file 20, and obtains the feature amount closest to the model for each data type. Identify the groups that you have. For this comparison processing, not only the method used in the first embodiment, but also correlation calculation and fuzzy inference may be used.

When the comparison process is completed for all the groups, the CPU 12 further retrieves the data regarding the character string from each specified group, recognizes it as the same kind of data as the corresponding model, and then the character string. Are stored in the database file 21.

FIG. 9 shows an example of the data structure of the database file 21. The data of each business card is extracted for each item and stored in the corresponding area. In this case,
The data of each business card is accumulated in the order of being read by the image scanner 7, but the data can be rearranged as necessary.

The operator operates the keyboard 11 to
When a search is performed using a specific data type as a key, the CPU 12
Retrieves the corresponding data from the database file 21 and outputs the result to the monitor 10 via the I / O interface 19. It should be noted that only specific data may be taken out from the database file 21 to create a new database file.

[0069]

As described above, according to the present invention, after all the characters included in a predetermined image are extracted, these characters are divided into groups based on the continuity of the characters, and the character string to be extracted is extracted from them. When the group corresponding to the model is determined and the information of the characters included in that group is output, when the characters other than the extraction target are included in the image or the positioning accuracy of the processing area is poor. However, the target character string can be extracted easily and accurately.

According to the second and fifth aspects of the invention, since the characters whose difference in the arrangement direction on the image is within a predetermined error range are classified into the same group, the character strings included in the image are classified. Grouping is realized.

In the inventions of claims 3 and 6, since the data relating to the character size is compared between each group and the character string model, the character string to be recognized is displaced or rotated. Even if it is, the target character string can be accurately extracted.

According to the seventh aspect of the invention, since the above method is applied to a plurality of character strings included in an image, each character string can be extracted easily and accurately.

According to the invention of claim 8, the images of the objects having a plurality of character strings are sequentially taken in, each character is extracted from the image of each object, the recognition processing is performed, and the result is stored. , It is possible to easily construct a database using the information of each object.

[Brief description of drawings]

FIG. 1 is a block diagram showing a circuit configuration of a character string extraction device according to an embodiment of the present invention.

FIG. 2 is an explanatory diagram showing a method for extracting character feature amount data.

FIG. 3 is an explanatory diagram showing a method of checking continuity of characters.

FIG. 4 is a flowchart showing a control procedure for extracting a character string.

FIG. 5 is an explanatory diagram showing a state in which characteristic amount data of all extracted characters are stored.

FIG. 6 is an explanatory diagram showing a state in which characteristic amount data of all set groups are stored.

FIG. 7 is a block diagram showing a circuit configuration of a database system using the character string extracting method of the present invention.

FIG. 8 is an explanatory diagram showing a data structure of a model data file.

FIG. 9 is an explanatory diagram showing a data structure of a database file.

[Explanation of symbols]

 1 model image data setting unit 2 image cutout unit 3 character extraction unit 4 classification processing unit 5 comparison processing unit 6 output processing unit 8 control device 9 magnetic disk file

 ─────────────────────────────────────────────────── ─── Continuation of the front page (72) Inventor Motoki Yotsui 10 Ouron Co., Ltd. 10 Hanazono Dodo-cho, Ukyo-ku, Kyoto City, Kyoto Prefecture

Claims (8)

[Claims] 1. A method for extracting a specific character string from an image containing a plurality of characters, the first step of extracting all the characters in the image, and extracting each extracted character. After the second step of grouping into character string groups having continuity, and comparing the feature quantity of each group with the model of the character quantity of the character string registered in advance, to determine the group to be extracted And a third step of outputting the information of the characters included in the group as the extraction result, in series. 2. The character string according to claim 1, wherein in the second step, characters whose difference in arrangement direction on an image is within a predetermined error range are classified into the same group. Extraction method. 3. The character string extraction method according to claim 1, wherein the feature amount compared in the third step is data relating to the size of a character included in the character string. 4. An apparatus for extracting a specific character string from an image including a plurality of characters, comprising model storage means for storing a model of a characteristic amount of the character string to be extracted, and the model storage means on the image. Character extraction means for setting a predetermined processing area and extracting all the characters contained in this area, and classification means for grouping the characters extracted by the character extraction means into groups for each character string having continuity A feature amount extraction means for extracting a feature amount for each set group, and a feature amount model of a character string stored in the model storage means for the feature amount of each group extracted by the feature amount extraction means. A character string extracting unit that includes a discriminating unit that discriminates a group corresponding to the model and an output unit that outputs information of characters in the group discriminated by the discriminating unit as an extraction result. Output device. 5. The classifying means is means for classifying characters having a difference in arrangement direction on an image in the processing area within a predetermined error range into the same group. Character string extraction device. 6. The character string extraction device according to claim 4, wherein the determination unit compares data relating to the size of characters included in the character string between each group and the model of the character string. 7. A device for extracting each character string from an image of an object in which a plurality of character strings are written and performing a predetermined recognition process, wherein a model of a feature amount of each character string to be extracted. A model storage means for storing a character, a character extraction means for setting a predetermined processing area on an image obtained by imaging an object and extracting all characters included in this area, and the character extraction means. A classifying unit that groups the extracted characters into groups for each character string having continuity, a feature amount extracting unit that extracts a feature amount of each set group, and each of the feature amounts extracting unit. Discriminating means for discriminating the group corresponding to each model by comparing the feature amount of the group with the model of the feature amount of each character string stored in the model storage means, and the character of each group discriminated by the discriminating means. And a recognition processing means for performing recognition processing by extracting the information of each of the above. 8. A system for extracting each character string from each of a plurality of objects on which a plurality of character strings are written, and performing a predetermined recognition process on each extracted character string, Image input means for sequentially capturing images of the target object, model storage means for storing a model of the characteristic amount of each character string to be extracted, and a predetermined processing area on the image of the target object captured by the image input means. Is set, and character extraction means for extracting all the characters contained in this area, and classification means for grouping the characters extracted by the character extraction means into groups for each character string having continuity are set. And a feature quantity extraction means for extracting a feature quantity for each group, and a feature quantity model of each character string stored in the model storage means for the feature quantity of each group extracted by the feature quantity extraction means, In comparison, a discriminating means for discriminating the group corresponding to each model, a recognition processing means for extracting and recognizing character information of each group discriminated by the discriminating means, and a processing result by the recognition processing means A character string recognition processing system comprising: storage means for storing the object for each object; and data output means for reading out the data stored in the storage means in a form according to a predetermined output instruction.