JP3074691B2 - Character recognition device - Google Patents

Description

Description: TECHNICAL FIELD The present invention relates to a character recognition device that recognizes a recognition target character by performing template matching using a plurality of templates.

<Conventional technology> Conventionally, in this type of character recognition apparatus, recognition is performed on a character to be recognized by focusing only on a character region of a template and determining a degree of coincidence between the character region and an image of the character to be recognized. The degree of matching is defined by the ratio of the number of matching pixels between the black pixels forming the character region of the template and the image of the recognition target character and the total number of black pixels forming the character region of the template.

<Problem to be Solved by the Invention> However, when the character to be recognized is an English character “h” as shown in FIG. 4 (3), the image 4 of the character to be recognized and the English character “h” are used. 4 (1), and the matching degree obtained by performing template matching with the template 1 (shown in FIG. 4 (1)) and the image 1 of the recognition target character and the template 1 (FIG. 4 (2) )), The matching degree obtained by performing template matching between the target character and the matching target character matches, so that it is possible to identify whether the image 4 of the recognition target character is an English character “h” or “n”. Not be recognized or misrecognized later.

In the figure, reference numeral 2 denotes a black pixel portion forming the character area of the template 1, and reference numeral 3 denotes a white pixel portion forming the background. Reference numeral 5 denotes a black pixel portion that is a constituent pixel of the recognition target character, and reference numeral 6 denotes a white pixel portion that forms the background.

The same applies to the case where the recognition target character is the English character "e" as shown in FIG. 5 (3), using the template 1 of the English character "e" (shown in FIG. 5 (1)). Since the matching degree obtained by performing the template matching and the matching degree obtained by performing the template matching using the template 1 (FIG. 5 (2)) of the English character “c” match, the recognition target character It is also difficult to identify the image 4.

In order to solve such a problem, the number of matching pixels between the white pixels 3 forming the background of the template 1 and the image 4 of the recognition target character, and the white pixels 3 forming the background of the template 1
The degree of coincidence with respect to the background is defined based on the ratio with respect to the total number of white pixels, and the character to be recognized is identified by comparing the degree of coincidence with the background.

However, in this method, noise 7 or underlined portion 8 is present in the image 4 of the recognition target character as shown in FIG. 6 (1), or the image of the recognition target character is displayed as shown in FIG. 6 (2). If part of the image of adjacent characters is included in 4,
Even if the degree of coincidence for the background is used, it is difficult to identify the character to be recognized.

The present invention reliably recognizes a character to be recognized even if the character to be recognized is any character whose form is similar to each other, and even if an image of the character to be recognized includes an image of noise or an adjacent character. It is an object to provide a character recognition device that can be identified. Another object of the present invention is as follows.
The point is that the identification process of the character to be recognized is speeded up by narrowing down the candidates for the character to be recognized from each template in a short time and performing the identification process.

<Means for Solving the Problems> A character recognition apparatus according to the present invention includes a storage unit for storing a plurality of templates, and sequentially reads out each template from the storage unit, and configures a character area of each template. Matching means for calculating the ratio of the number of matching pixels between the pixels constituting the character region of the template and the constituent pixels of the recognition target character with respect to the number of pixels as the degree of matching between the character region of the template and the image of the recognition target character,
Detecting means for comparing the degree of matching calculated by the matching means with a threshold value and detecting a character of the template having a degree of matching exceeding the threshold value as a candidate for a character to be recognized; Identification means for identifying, when a candidate character is detected, a candidate character of the template in which the number of matching pixels between the pixels constituting the character region of the template and the constituent pixels of the recognition target character is the largest, as the recognition target character. Become.

<Operation> After detecting a candidate for a recognition target character based on the degree of matching between the image of the recognition target character and the character area of each template, when a plurality of candidate characters are detected, one of the candidate characters is referred to by referring to the number of matching pixels. Is identified as a recognition target character, so that the recognition target character can be distinguished from any other character whose form is similar to each other. In addition, since the degree of matching between the character region of the template and the image of the character to be recognized is calculated by focusing only on the character region of the template, even if the image of the character to be recognized contains noise, an image of an adjacent character, or the like. Have no adverse effect on character identification. Furthermore, a template as a candidate for a recognition target character is detected based on the ratio of the number of pixels corresponding to the constituent pixels of the recognition target character in the character area of the template. , The speed of the identification process of the recognition target character can be increased.

<Embodiment> FIG. 1 is a character recognition device according to an embodiment of the present invention.
11 shows the entire configuration.

The device 11 in the illustrated example is of a type in which a template is entirely scanned with respect to an input image of a character to be recognized, and an image input device 12, an image memory 13, a matching processing unit 14, an identification unit 23,
A microprocessor unit (hereinafter abbreviated as MPU) 15 and the like are included as components, and an operation unit 16 such as a keyboard and a host computer 17 are connected to the MPU 15 via an external interface circuit 18.

The image input device 12 fetches a reference image of a template to be registered in the teaching mode and stores it in the image memory 13, inputs an image of a recognition target character in the recognition mode, and provides the image data to the matching processing unit 14.

The reference image stored in the image memory 13 is read out by the MPU 15, and the image of each character whose area is specified by the operation unit 16 is registered in the pattern memory 22 as a template.

The matching processing unit 14 includes a pattern matching unit 20,
It includes a coincidence degree calculation unit 21 and a coincidence point detection unit 24,
Image data for one frame of the character to be recognized is input from the image input device 12 to the pattern matching unit 20.
A template is loaded by U15, and template matching is performed by sequentially scanning the entire template with respect to this input image.

The matching degree calculation unit 21 calculates, as a matching degree, a ratio of the number of matching pixels between pixels (black pixels) constituting the character region of the template and the input image and the total number of black pixels constituting the character region of the template. It is. The coincidence point detection unit 24 compares the degree of coincidence calculated by the degree of coincidence calculation unit 21 with a predetermined threshold value given by the MPU 15, and determines a template character having a degree of coincidence exceeding the threshold value as a recognition target character. It is detected as a candidate, and the matching position is output to the MPU 15 as the position of the recognition target character. The coincidence point detection unit 24 stores the number of coincident pixels between the black pixels constituting the character area of the template and the input image for each coincidence position in a predetermined storage area.

The identification unit 23 is a unit that, when a plurality of candidate characters are detected at the character position detected by the matching point detection unit 24, identifies any one of the candidate characters as a recognition target character. In the identification unit 23, candidate data from the MPU 15 and the number of matching pixels obtained at the time of extraction of each candidate from the matching point detection unit 24 are input, and candidate characters having the maximum number of matching pixels are set as recognition target characters. Identify.

FIG. 2 shows an operation procedure of the character recognition device 11 in the recognition mode.

First, prior to template matching, the MPU 15 reads the first template from the pattern memory 22 and scans the inside of the template in steps 1 (indicated by “ST1” in the figure) to step 3 of FIG. The total number of black pixels to be processed is counted. In Step 1, it is determined whether or not the first pixel in the template is a constituent pixel (black pixel) in the character area, and the determination value of "YES" increments the count value of the total number of black pixels (Step 2). When the same determination and counting operation is performed for all the pixels in the template, step 3 is “YES”, and the flow shifts to the next processing.

At this stage, the image input device 12 supplies the image data of the recognition target character to the pattern matching unit 20 of the matching processing unit 14, and the MPU 15 loads the first template into the matching processing unit 14, and the input image is processed according to this template. Template matching is performed.

In this template matching, in step 4, it is determined whether or not the first black pixel forming the character area of the template matches the input image, and if the determination is "YES", the count value of the number of matching pixels is incremented ( Step 5). When the same determination and counting operation is performed by the matching degree detection unit 21 for all the pixels constituting the character area in the template, the determination in step 6 becomes “YES”,
In the next step 7, the ratio between the count value obtained in step 5 and the count value obtained in step 2 is calculated as the degree of coincidence.

Next, in step 8, the coincidence point detection unit 24 determines whether or not the calculated value of the degree of coincidence exceeds a predetermined threshold value. If the determination is “YES”, the matching point of the current template is determined. It is determined that there is a possibility that the character of the template exists at the scanning position, and the MPU 15 takes in the coordinates of the current position and stores the character as a candidate character.

In step 10, it is determined whether or not the entire scanning of the input image by the template has been completed. If the determination is "NO", the process proceeds to the next scanning position and the same procedure as described above is executed. .

Thus, if the determination in step 10 is "YES", template matching using the next template is executed in a similar procedure.

When the matching process using all the templates is completed, step 11 becomes “YES” and the process proceeds to step 12, where the integration process is performed on the candidate group of each character position to identify each character position. When the integration process is completed for the candidate group at each character position, step 13 becomes "YES", and in the next step 14, it is determined whether or not a plurality of candidate characters are detected at the first character position.

If the determination in step 14 is “YES”, the identification unit 23
Identifies in step 15 the candidate character having the maximum value among the number of matching pixels obtained in step 5 for each candidate character as the recognition target character. If a plurality of candidate characters are not detected at the focused character position, the determination in step 14 is “NO”, and the single candidate character is identified as a recognition target character.

Thus, when the same identification processing is performed for all character positions, step 16 becomes "YES" and the character recognition procedure is completed.

For example, if the character to be recognized is the English letter “h” shown in FIG. 4 (3), the image 4 and the respective templates of the English letters “h” and “n” shown in FIG. 4 (1) (2) When template matching is performed for each of them, the matching degree obtained in step 7 is common to both. However, since the number of matching pixels obtained in step 5 is larger for the template “h”, the character to be recognized is identified as “h”.

The same applies to the case where the recognition target character is the English character “e” shown in FIG. 5 (3), and the number of matching pixels obtained in step 5 is larger in the template “e” than in the template “c”. Therefore, the recognition target character is identified as “e”.

Also, since the degree of coincidence is calculated by focusing only on the character region of the template, even if the input image has noise 7 or underlined portion 8 as shown in FIG. Thus, highly accurate character recognition can be performed without any trouble.

FIG. 3 shows a configuration of a character recognition device 11a according to another embodiment of the present invention. In this embodiment, after the input image is stored in the image memory 13, each character area of the input image is specified by the area specifying unit 25, and one character is cut out by the character cutout unit 19. Then, template matching is performed on the extracted image of the recognition target character,
The processing of the above-described embodiment is executed.

<Effect of the Invention> As described above, the present invention detects a candidate for a recognition target character based on the degree of matching between the image of the recognition target character and the character area of each template, and then detects a matching pixel when a plurality of candidate characters are detected. Since any candidate character is identified as a recognition target character by referring to the number, even if the recognition target character is any character whose shape is similar to each other, it can be distinguished. Even if noise or an image of an adjacent character is included in the image, the character identification is not affected at all and character recognition can be performed reliably and with high accuracy. Further, prior to the identification process, the candidates of the recognition target character are narrowed down by using the ratio of the number of pixels corresponding to the constituent pixels of the recognition target character in the character region of the template. Thus, the character to be recognized can be identified with high accuracy and high speed.

[Brief description of the drawings]

FIG. 1 is a block diagram of a character recognition apparatus according to one embodiment of the present invention, FIG. 2 is a flowchart showing a procedure of character recognition, FIG. 3 is a block diagram showing a configuration of another embodiment of the present invention, and FIG. 4 and 5 are explanatory diagrams showing a template and a character to be recognized in comparison with each other, and FIG. 6 is an explanatory diagram for explaining a problem of the prior art. 11, 11a: Character recognition device 14: Matching processing unit, 15: MPU 20: Pattern matching unit 21: Matching degree calculation unit 22: Pattern memory, 23: Identification unit 24: Matching point detection unit

──────────────────────────────────────────────────続 き Continued on front page (58) Field surveyed (Int.Cl. ⁷ , DB name) G06K 9/62

Claims (1)

(57) [Claims] 1. A storage unit for storing a plurality of templates, and each template is sequentially read out from the storage unit, and a pixel constituting a character area of the template with respect to a total number of pixels constituting a character area of the template. Matching means for calculating the ratio of the number of matching pixels between the image data of the character to be recognized and the image of the character to be recognized, and Detecting means for detecting a template character having a matching degree exceeding the threshold value as a candidate for a recognition target character, and detecting a character of the template when a plurality of candidate characters are detected by the detecting means. A template candidate character in which the number of matching pixels between the pixels constituting the region and the pixels constituting the recognition target character is the largest, Character recognition device formed by and a identification means for identifying a identification target character.