JPH0535921A - Device and method for collating character - Google Patents

Abstract

PURPOSE:To improve recognition accuracy by setting an optimum binarizing level for every picture. CONSTITUTION:A character to be discriminated is read (S1). The read picture is binarized by a first binarizing level (S2, S3). Characters are segmented from the binarized picture one by one (S4). The segment character is collated with a standard pattern (S5). Whether a coincidence degree exceeds a prescribed value or not is decided, and a first signal is outputted when it is decided that the degree exceeds and a second signal is outputted when it is decided that the degree does not exceed (S6). In response to the second signal, a second binarizing level relating to the first binarizing level is set (S9). Characters are segmented by the second binarizing level (S4). In response to the first signal, the next segmented character is collated (S5).

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a character collating apparatus by image processing and a method therefor, and particularly, it is suitable even when the colors of characters in the measurement area are different from each other or the illumination is uneven. BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a character matching device that can perform character matching and a method therefor.

In this specification, the word "character" is used in a broad sense to include not only characters such as hiragana, katakana and alphabets but also numbers and symbols.

[0003]

2. Description of the Related Art Characters marked on the surface of a member having a circular cross section (for example, a steel material such as a cylindrical coil) may include bleeding or blurring, and since it is drawn on a curved surface, it is uniform. Since it cannot be illuminated, it is important to detect and set the binarization level for reading the character and thus for determining the shade, in order to enable such automatic recognition of the character.

The conventional character recognizing apparatus and recognizing method have been premised on that the binarization is appropriately performed depending on the shade of the character to be read. As such an example, form a closed loop of a character line element, crossing the characteristic parameters of the endpoints such as compared to the reference characters character by character, and discloses a method and apparatus for performing character recognition by-nearest Patents
Japanese Patent Laid-Open No. 1-229388 , a conventional reference value in which pixel density (appearance frequency) of each character in the XY direction is patterned is compared with a target character, and when a character is determined by an approximate value, noise is generated. A pattern matching method for extracting a part and matching the remaining part excluding the noise part has been disclosed.
Japanese Unexamined Patent Publication ( Kokai) No. 2-68683 discloses a pixel group that forms a reference character pattern, a pixel group that is a non-character area for all the reference character patterns, and a character group that is a non-character area in the reference character pattern but is a non-character area Japanese Unexamined Patent Application Publication No. 2-68681, which discloses a character recognition method in which a pixel group that becomes a region and a pixel group that becomes a character region in the reference character pattern are separately weighted and the match rate is evaluated, can be cited. .

These character recognition devices and character recognition methods are
As mentioned above, it is premised that the binarization is performed correctly, and even if the separation and removal of noise etc. is performed, the main cause is another cause. Also, character bleeding, color change, shading It did not follow the changes in.

Moreover, in the conventional character recognition device,
Since the entire area for character recognition is binarized at a single and the same level, even if an attempt is made to recognize a character marked on a curved surface such as the outer surface of a cylindrical steel material by such a device, there is uneven illumination. It is often impossible to obtain a clear binary image from the entire area because of the influence of lens shading.

[0007]

Generally, when a plurality of areas are inspected, a binarization level is set for each area even if the shape, color, glossiness, etc. of the characters existing in each area are different. Need to be able to.

The present invention has been made in order to improve the conventional drawback of recognizing characters in the entire region with a single and same binarization level, and clear binarization can be obtained in the process of character collation. An object of the present invention is to provide a character collating apparatus and method for setting an optimum binarization level for each area when there is no such character.

[0009]

In order to achieve the above object, a character collating apparatus according to the present invention comprises an image reading means for reading a character to be identified and an image read by the reading means. Binarization means for binarizing at a first binarization level, cutting-out means for cutting out the character from the binarized image, outputting the cut-out characters one by one, and pre-cutting-out The collating means for collating the character output from the means with the standard pattern, and whether or not the degree of coincidence in the collation by the collating means exceeds a predetermined value is judged. And a first determining unit that outputs a second signal when it is determined that the first signal does not exceed the first signal,
The second binarization level associated with the binarization level of
The quantizing means is set and the clipping means is set to the second 2
The means for cutting out a character at a digitization level and the first
In response to the signal, the means for cutting out the next character from the cutting-out means.

The character collating method according to the present invention further comprises an image reading step for reading a character to be identified,
A binarization step for binarizing the image read in the reading step at a first binarization level; cutting out the character from the binarized image; outputting the cut out characters one by one And a matching step for matching the output character with a standard pattern, and whether the degree of matching in the matching in the matching step exceeds a predetermined value. A first signal, and a second signal is output when it is determined that the first signal has not been exceeded, and the first signal is output in response to the second signal.
Setting a second binarization level associated with the binarization level of, and cutting out characters at the second binarization level in the cutting out step, and the cutting out in response to the first signal. Outputting the next character cut out in step.

A character collating apparatus and a character collating method according to the present invention are used when a character is marked on a curved surface or when the character is a character string including a plurality of characters, and at least one of the characters has a color. Suitable when different from other characters.

The second binarization level is determined by the difference between the ratio of the character part of the character cut out and output and the ratio of the character part of the standard pattern matched with the character.
This is a modification of the valuation level.

[0013]

According to the present invention, it is determined whether or not the degree of coincidence in matching one image with a standard pattern exceeds a predetermined value, and if it does not exceed the predetermined value, the binarization level for cutting out the image is associated with it. To another binarization level, the image is cut out again with the changed binarization level, and the degree of coincidence at the time of matching with the standard pattern is determined again.

[0014]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the present invention will be described below in detail with reference to the drawings. FIG. 1 shows an actual usage state of the character collating apparatus according to the present invention. In the figure, a number string 2 for identifying the coil is marked on the curved outer surface of the coil 1. The numeral string 2 is illuminated by the lighting device 3, but since the numeral string 2 is marked on the curved surface, each area where the numeral is marked tends to be unevenly illuminated. The character matching device 4 according to the present invention is arranged so as to face the numeral string 2. Character matching device 4
Is a camera for reading the numbers in the number string 2, an A / D converter for A / D converting the output of the camera, a first memory for storing the output of the A / D converter, and a standard character, as in the conventional device. A second memory for storing a pattern, a third memory for storing a collation result, and outputs of the first and second memories are compared to collate a plurality of numbers one by one, and the result is stored in a third memory. A processor to be stored in, and
A display device for displaying the result of the collation is included. Character matching device 4
Is connected to a higher-order computer 5, information from the higher-order computer 5 to the character collating device 4 is transmitted to the character collating device 4 (for example, coil number), and a standard pattern for the number sequence is sent based on the information. May be determined.

Next, the operation of the character collating apparatus according to the present invention will be described with reference to the flowchart of FIG.

First, in step S1, the camera of the character collating device 4 captures the number string 2 as an object. Subsequently, in step S2, the binarization level X of the captured image is obtained by the density histogram measurement,
In step S3, the image of the number string 2 is binarized by using this binarization level X (binarization at the same level in all areas).

Next, in step S4, an image corresponding to each number is cut out. For this,
The lateral distribution of the pixels in the lateral direction (the value of the pixels is summed up one by one in the lateral direction) is determined, and the position of each image is determined.
Then, the vertical distribution of the pixels is obtained, and each image is cut out. In the image cut out in this manner, the variation in the image size due to the difference in the distance between each number and the camera is corrected (normalization of the number size).

Although the respective numbers forming the number string are marked at a substantially constant pitch, it is common that the positions of the marked numbers are vertically displaced or the sizes of the numbers are not uniform. Therefore, when importing a number string, set a blank corresponding to a certain number of pixels above and below the number string to determine the reference reading height, and according to that, determine the reading width of one image to determine the reading width of each image. Find the center position. By this, the optimum reading center position is decided,
It suffices to cut out one number from each end of the number string at a predetermined pitch.

Subsequently, in step S5, the one image cut out in step S4 is collated with the standard pattern stored in the second memory by pattern matching, and the degree of coincidence in the image is calculated.

A step S6 checks whether or not the degree of coincidence in the collation in the step S5 exceeds a predetermined value (for example, 90%). If the degree of coincidence exceeds a predetermined value, it is determined that proper matching has been performed, the process proceeds to step S7, and it is determined whether the matching of all the binarized images in step S3 is completed. If all images have not been collated, the process returns to step S5, and the next image obtained at the binarization level X is collated with the standard pattern in step S3, and then all images are collated in the same manner. Go.

However, if the degree of coincidence in the collation in step S5 is less than the predetermined value, that is, if the erroneous collation state is confirmed in step S6, the binarization level is changed a predetermined number of times, and each time the image concerned is changed. It is determined whether the degree of coincidence between the standard pattern and the standard pattern exceeds a predetermined value. That is, when it is determined in step S6 that the degree of coincidence does not exceed the predetermined value, the process proceeds to step S8, and it is determined whether or not the occurrence of the erroneous collation state has reached a predetermined number (for example, twice). In this case, it is first determined that the collation state is incorrect, and thus the process proceeds to step S9.

In step S9, step S4
The ratio of the number of pixels of the image cut out in step 1 to the total number of pixels of the area including the image, that is, the ratio (A) of the character portion is calculated,
The ratio (A) is compared with the ratio of the number of pixels of the number of the standard pattern determined in step S4 to the total number of pixels of the area including the standard pattern, that is, the ratio of the character portion (B),
A new binarization level Y (= X + kC; k is a constant) is set based on the difference (C = A to B), and step S3 is performed.
Return to.

Therefore, in step S3, the reset 2
The image is binarized based on the binarization level Y, the image is cut out again, and the image cut out in this way and the standard pattern are collated in step S5. In step S6, it is determined whether the degree of coincidence at that time exceeds a predetermined value,
If it is confirmed that the degree of coincidence exceeds the predetermined value, it is determined that proper collation has been performed, and therefore, step S7.
Proceeding to step S5, the next image (which is the image first cut out at the binarization level X in step S3) is displayed in step S5.
In S6 and S7, the pattern is compared with the standard pattern and processed. Also in this case, if it is determined in step S6 that the degree of coincidence does not exceed the predetermined value, the new binarization level Z is reset in association with the previous binarization level by the method described above. If the degree of coincidence exceeds a predetermined value before the binarization level is reset for more than a predetermined number of times, the binarization level at that time is optimized. It is determined that it is a digitization level, and it is determined that proper matching has been performed. In this way, the optimum binarization level is set for each image, and the matching of all images is performed based on that.

However, if the degree of coincidence does not exceed the predetermined value even if the binarization level is reset in step S9 a predetermined number of times, step S8 determines that the collation state is a predetermined number of times. Check that, it is judged that the material is erroneously input, and the work is finished.

FIG. 3 shows an image in which the numeral 6 marked on the curved surface is binarized by the conventional device and an image when binarized by the character collating device according to the present invention. As can be seen from this figure, the image binarized according to the present invention is sharper.

Although one embodiment of the present invention has been described in detail above, the present invention is not limited to the above embodiment. For example, in the above-described embodiment, the image is binarized in step S3 with a single and the same binarization level in the entire area. Instead, however, the image of the number string captured in step S1 is cut out. It is also possible to perform a process to detect a measurement region of each image and set a binarization level individually for each region to obtain a binarized image.

Further, although the above-mentioned embodiment relates to the processing when the numeral string is marked on the curved surface, it can be applied to the case where the numeral string includes numerals of different colors or the illumination is uneven. You can

In this way, also when collating a number string marked on a curved surface such as a coil, the binarization level is set individually and in an analog manner for each image to form a binarized image. Since it is obtained and the character collation is performed based on this, it is possible to reduce erroneous recognition due to noise due to number bleeding, blurring, uneven lighting, etc. Therefore, it is possible to prevent and confirm erroneous coil processing. The necessary personnel can be omitted.

Further, even when the colors for marking the numbers are different or the illumination of the characters is uneven, the optimum binarization level can be set for each image, so that the character recognition accuracy is improved. Therefore, it is possible to expand the field to which the automatic recognition of characters is applied and flexibly deal with various cases.

[0030]

As is apparent from the above description of the present invention in detail with reference to the examples, when characters are marked on a curved surface, when characters marked in different colors are recognized, the characters are uniform. If the matching degree of character matching does not exceed the specified value, such as when the
Since the binarization level is changed a predetermined number of times in association with the previous binarization level, and each time the character is collated with the changed binarization level, the collation accuracy can be improved, and therefore, Not only can the automatic recognition of characters be applied to a wider range, but the number of personnel for preventing erroneous processing due to erroneous recognition of characters can be reduced.

[Brief description of drawings]

FIG. 1 is a diagram illustrating a usage state of a character matching device according to the present invention.

FIG. 2 is a flowchart showing the operation of the character collating device according to the present invention.

3A is a diagram showing a binarized image by a conventional device, and FIG. 3B is a diagram showing an image when binarized by changing the binarization level according to the character collating device according to the present invention.

[Explanation of symbols]

1 coil 2 Number sequence 3 lighting equipment 4-character collation device 5 Host computer

Claims (8)

[Claims] 1. An image reading unit for reading a character to be identified; a binarizing unit for binarizing an image read by the reading unit at a first binarizing level; A cutting-out means for cutting out the characters one by one from the binarized image, a matching means for matching the characters cut out from the front cutting-out means with a standard pattern, and a matching degree in matching by the matching means is predetermined. It is determined whether or not the value has been exceeded. If it is determined that the value has been exceeded, a first signal is output, and if it is determined that the value has not been exceeded, a second signal is output, and in response to the second signal. And means for causing the binarizing means to set a second binarizing level associated with the first binarizing level, and causing the clipping means to clip a character at the second binarizing level. And the first signal To the character collation apparatus and means for causing cut the next character from the cut-out means. 2. The character matching device according to claim 1, wherein the character is marked on a curved surface. 3. The character collating apparatus according to claim 1, wherein the character is a character string including a plurality of characters, and at least one of the characters has a color different from that of the other characters. 4. The first binarization level is determined by the difference between the ratio of the character portion of the character output in the cutting step and the ratio of the character portion of the standard pattern matched with the character. The character collating device according to claim 2, wherein the binarization level is corrected. 5. An image reading step of reading a character to be identified; a binarizing step of binarizing an image read in the reading step at a first binarizing level; and the binarizing step. A step of cutting out the characters from the image one by one, a matching step of matching the cut-out characters with a standard pattern, and determining whether the degree of matching in the matching in the matching step exceeds a predetermined value, and And a second signal when the first signal is not exceeded, and a second signal when the first signal is not exceeded, and the first binarization level in response to the second signal. A second binarization level associated with the step of cutting out characters at the second binarization level in the cutting out step, and in the cutting out step in response to the first signal. Next character Character collation method comprising the steps of force. 6. The character matching method according to claim 5, wherein the character is marked on a curved surface. 7. The character matching method according to claim 5, wherein the character is a character string including a plurality of characters, and at least one of the characters has a color different from that of the other characters. 8. The second binarization level is the first binarization level based on a difference between a ratio of a character portion of a character output in the clipping step and a ratio of a character portion of a standard pattern matched with the character. 8. The character matching method according to claim 6, wherein the binarization level is modified.