JP2972516B2 - Print pattern inspection device for inkjet printer - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an apparatus for inspecting a printing pattern of an ink jet printer, and more particularly to an apparatus for inspecting and recognizing a printing pattern printed by an ink jet method on each work conveyed by a conveyor or the like.

[0002]

2. Description of the Related Art In a conventional print pattern recognition / inspection apparatus, preprocessing (noise removal processing and the like) and pattern recognition processing generally used in image recognition technology are generally employed.

For example, in the case of performing a recognition inspection of a pattern printed on a work to be printed by an ink jet method, the print pattern is imaged by a camera, noise is removed by preprocessing, and a preset pattern is set. Pattern recognition and character recognition are performed by extracting a printing dot pattern at the binarization level and comparing it with a reference dot pattern.

[0004]

In such a system using pre-processing and pattern recognition in the conventional image recognition technology, as shown in FIG. 8A, a sample to be inspected in which each dot is clearly divided is shown. Shows a good recognition rate, but as shown in FIG. 8B, for a sample in which each dot is connected and is not clearly divided, the sample is compared with a reference pattern in which each dot is clearly divided. For this reason, there are disadvantages that the recognition rate is remarkably lowered and that excessive printing defects are detected.

In particular, in the case of an ink jet printer which performs printing while moving each work such as cans and bins by a conveyor device such as a conveyor, each dot constituting a print pattern is formed as shown in FIG. As shown in FIG. 8 (B), there are many drawbacks in that the conventional methods do not provide a sufficient recognition rate and do not provide a proper pass / fail judgment rate.

SUMMARY OF THE INVENTION An object of the present invention is to provide an ink jet printer capable of recognizing a printing pattern in which a large number of connections between dots occur due to fluctuations in a time axis direction which is a printing direction and division of each dot is not clear, with high accuracy. An object of the present invention is to provide a print pattern inspection apparatus.

[0007]

A print pattern inspection apparatus according to the present invention regards a reading means for reading a print pattern by an ink jet system and a printing direction of an image read by the reading means as a time axis of a formant in a voice recognition technique. Further, a pre-processing means for obtaining a time-series feature pattern corresponding to a formant in a speech recognition technology by regarding a print pattern height direction of the read image as a frequency axis of the formant, and the time-series feature pattern as a reference pattern. And a recognition unit that performs a recognition process while comparing with a reference time-sequentialized feature pattern.

[0008]

Embodiments of the present invention will be described below with reference to the drawings.

FIG. 1 is a schematic block diagram of an embodiment of the present invention. A work 3 such as a can or a bottle is conveyed at a predetermined speed in the X direction by a conveyor 1 which is a conveying device, and desired printing is performed on each of these works 3 by an ink jet printer 2. .

The print pattern printed on each of these works 3 is read by an imaging camera 4, derived as an image 5, and supplied to a preprocessing unit 6. In the pre-processing unit 6, as will be described in detail later, binarization (or multi-valued as necessary)
After undergoing noise removal, correction in the height direction of the print pattern (y direction: see FIG. 2B), and extraction of the print character area, the time-series feature pattern 7 is generated.

FIG. 2 shows the time series characteristic pattern 7.
This will be described with reference to FIG. FIG. 2A shows an example of a formant generally used in the speech recognition technology, and a temporal change of a frequency spectrum of the speech is displayed with a horizontal axis representing a time axis t and a vertical axis representing a frequency axis f. It was done.

In the present invention, a time-sequential feature pattern 7 of a print pattern corresponding to a formant in the speech recognition technique is obtained by the pre-processing unit 6 as shown in FIG.
The next recognition inspection unit 8 makes a pass / fail judgment by comparing with a standard time-series feature pattern.

The time-series characteristic pattern of the print pattern obtained from the preprocessing section 6 has a print direction x corresponding to the time axis t direction of the formant in FIG. 2A, as shown in FIG. The pattern is such that the height direction y corresponds to the direction of the frequency axis f of the formant. The time-series feature pattern 7 is compared with the time-series feature pattern of the printing standard pattern stored in the standard pattern file 9 in advance.

FIG. 3 is a flowchart showing the operation processing of the preprocessing unit 6. The read image picked up by the camera 4 is subjected to a binarization or multi-value processing in step 31. Here, the binarization process is to compare each pixel of the read image with a threshold value and convert it into 1-bit binary data of “1” and “0” according to the magnitude. Further, the multi-value processing is to convert the luminance of each pixel of the read image into, for example, 256-level information. In this case, the luminance is converted into 8-bit data. Whether to perform the binarization processing or the multi-value processing is selected according to the system.

Next, at step 32, noise removal is performed. As an example of a noise removing method, there is a well-known technique of dilation, contraction, and smoothing. Dilation and contraction are applied to binary image data, and smoothing is applied to multilevel image data. is there.

Expansion is an operation of replacing eight neighborhoods (eight adjacent pixels) of white pixels in a binary image with white pixels, and contraction is an operation of replacing eight neighborhoods (neighboring eight pixels) of black pixels in a binary image. Pixel) is replaced with a black pixel.

By performing the deflation operation after performing the expansion operation and conversely, performing the expansion operation after performing the deflation operation, independent point-like noise can be removed.

Smoothing is an operation in which an output value of a pixel at coordinates (i, j) is set to an average of density values of pixels in a local area centered on a pixel at coordinates (i, j) in the original image. This is a noise removal technique for multi-valued images.

Next, the tilt correction processing in the y direction is performed in step 3
3 is performed. This inclination correction in the y-direction is performed by correcting the y-direction of the printed pattern caused by the fluctuation in the y-direction of the conveyor.
FIG. 4 shows a detailed flowchart of the correction process.

In the case of a multi-valued image, a binarization process is performed on the image from which noise has been removed in step 41 according to a specified value, and in the case of a binarized image, the binarization process is not performed.

In step 42, the position of the center of gravity of the character in the character window set so as to include the first character is calculated and is set to (g1x, g1y). Then step 4
In step 3, the position of the center of gravity of the character in the character window set to include the n-th character (the last character) is calculated, and (g
nx, gny).

In step 44, the inclination α is obtained as α = (gny-g1y) / (gnx-g1x).

In steps 45 and 46, the center coordinates of the character window of the i-th character are set to (Wix, Wi
As y), the correction amount di in the y direction of the i-th character is calculated as di = α (Wix−g1x) −Wiy.

By changing i from 2 to n-1 (for the 2nd to n-1st characters), the data in the i-th character window is shifted by di in the y direction to perform tilt correction ( Steps 45-48).

Here, the first (i = 1) and the last (i
= N), the center of gravity is calculated for the two characters. However, the present invention is not limited to these characters where i = 1 and n, and can be applied to any character.

Returning to FIG. 3, in steps 34 and 35, a process of extracting a print character area of a print pattern is performed. First, in step 34, projection data of the image in the y-axis direction is created as shown in FIG. And step 35
It is determined that a portion of the projection data which is equal to or greater than the predetermined threshold value Vt is the character region 40, and this portion is derived as a time-series feature pattern in step 36, and the recognition inspection unit 8 shown in FIG.
Output to

Next, a method for recognizing and inspecting the time-series feature pattern 6 in the recognizing and inspecting unit 8 will be described with reference to the drawings. FIG. 6 is a flowchart showing a procedure for performing recognition and inspection on the time-series feature pattern, and FIG. 7 is an explanatory diagram thereof.

A pointer i corresponding to the x-axis coordinate which is the printing direction of the time-series feature pattern 61 extracted by the preprocessing unit 6 and an x-axis coordinate of the reference data 62 registered in the standard pattern file 9 in advance. The pointer j to be executed and the error counter e are initialized to "0" (step 6).
1).

K is changed from 0 to N which is defined in advance as a search range, and the degree of mismatch fi + k, j of the pattern data is calculated by the following equation (step 62). As described later, N is suitably a value of 0 to 2.

Fi + k, j = | xi + k-mj | where xi + k is the pattern data in the character height direction y when the printing direction x-axis of the time-series feature pattern 6 is the (i + k) th axis;
mj is pattern data in the character height direction y when the print direction x-axis of the reference data 15 registered in advance is j-th.

Next, when k is changed from 0 to N, fi +
Find the value of k at which fi + k, j is the smallest among k, j and call it k1. Add fi + k1, j to the error counter e, i
Is added to k1 and j is added to 1 (steps 63 and 64).
The above operation is repeated until i or j has finished pointing to the final data (step 65).

If the value of the error counter e at this time is within the allowable value, the reference pattern 15 and the time-series feature pattern 6
Are determined to be “good”, and otherwise determined to be “bad” (steps 66 to 68).

For example, assuming that the search range N is 1, the time-series feature pattern 61 shown in FIG. 7 is associated with the reference pattern 62 as indicated by arrows a1 to a9. As can be seen from the example of FIG. 7, the value of N is determined in consideration of the pattern distortion caused by the displacement of the print pattern 61 in the x direction (time axis direction). 0 to 2 is appropriate, and if it is larger than that, there is a possibility that comparison with adjacent character patterns may be made, and N = 1 is optimal as shown in FIG.

As described above, it is possible to inspect whether the captured image matches the reference pattern, that is, whether the image is good or defective.

In the above description, the case of inspecting an image has been described, but a plurality of reference patterns are provided.
It is also possible to perform image recognition by calculating an error counter e for each reference pattern and searching for which reference pattern makes the error counter e the smallest.

In general, in the case of printing by an ink jet printer, the intervals between printed dots become irregular as shown in FIG. 8B due to uneven speed of a conveyor, clogging of nozzles of the ink jet printer, or the like. Dots are often connected. In the conventional method, such a print pattern is recognized by the reference pattern and the pattern recognition, so that the recognition rate is naturally low.

On the other hand, in the speech recognition technique, as shown in FIG. 2, recognition is performed based on the distribution state of the formant frequency f of the speech on the time axis t. Stable recognition taking into account the change in direction is possible, and the recognition technology has been established.

Therefore, in the present invention, the character height direction y of the image is regarded as the frequency axis f in speech recognition, and the image printing direction x is also regarded as the time axis t. The above-described established speech recognition technology is applied, and an accurate recognition test of a printed pattern is performed.

In the above description, the case where a two-dimensional camera is used as the camera 4 has been described. However, a one-dimensional camera can be used. In this case, the printing direction x becomes the time axis without any change. The time series in step 6 becomes unnecessary.

[0040]

As described above, according to the present invention, the recognition technique based on the formant used in the voice recognition technique is employed for the recognition inspection of the print pattern of the ink jet printer, so that the printing direction and the character height direction are used. There is an effect that a high recognition rate can be obtained even with respect to misalignment.

[Brief description of the drawings]

FIG. 1 is a block diagram of an embodiment of the present invention.

FIG. 2A is a diagram illustrating an example of a formant used in a speech recognition technique, and FIG. 2B is a diagram illustrating an example of a print pattern corresponding to the formant.

FIG. 3 is a flowchart showing an operation of a preprocessing unit 6 in FIG. 1;

FIG. 4 is a flowchart showing details of y-direction tilt correction (step 33) of FIG. 3;

FIG. 5 is a diagram for explaining the processing of steps 34 and 35 in FIG. 3;

FIG. 6 is a flowchart showing the operation of the recognition inspection unit 8 of FIG. 1;

FIG. 7 is a diagram illustrating an example of a recognition test performed by a recognition test unit 8;

FIG. 8A is a diagram showing an example of a good print pattern of the ink jet printer, and FIG. 8B is a diagram showing an example of the same disorder.

[Explanation of symbols]

 Reference Signs List 1 Conveyor 2 Inkjet printer 3 Work 4 Camera 6 Preprocessing unit 8 Recognition inspection unit 9 Inspection pattern file

Continuation of the front page (72) Inventor Masanori Kimura 5-7-1 Shiba, Minato-ku, Tokyo NEC Corporation (72) Inventor Midori 5-7-1 Shiba, Minato-ku, Tokyo NEC Corporation (72) Inventor Takeshi Fukuhara 5-8-20 Mukojima, Sumida-ku, Tokyo Asahi Breweries Engineering Tokyo Co., Ltd. (58) Field surveyed (Int.Cl. ⁶ , DB name) G01N 21/88 G01N 21/89 B41J 2/01 G06T 7/00

Claims (3)

(57) [Claims] 1. A reading means for reading a print pattern by an ink jet system, a printing direction of an image read by the reading means is regarded as a time axis of a formant in a speech recognition technology, and a height direction of a printing pattern of the read image is regarded as the formant. Preprocessing means for obtaining a time-series feature pattern corresponding to a formant in speech recognition technology assuming that the frequency axis is a frequency axis, and performing recognition processing while comparing this time-series feature pattern with a reference time-series feature pattern which is a reference pattern. A printing pattern inspection apparatus for an ink jet printer, comprising: 2. The method according to claim 1, wherein the recognizing unit compares a time-series feature pattern of the read image with a time-based feature pattern and a feature pattern of the reference time-series feature pattern at each corresponding time with each other. 2. A print pattern inspection apparatus for an ink jet printer according to claim 1, wherein said print pattern inspection apparatus is configured to judge whether or not the print pattern is good or bad. 3. The pre-processing unit includes: a noise removing unit that removes noise from the read image; a correcting unit that performs a correction process in a height direction of the print pattern on the image after the noise removal; 3. An ink-jet apparatus according to claim 1, further comprising: an extracting unit for extracting a print character area from a subsequent image, wherein the time-series feature pattern is obtained from the extracted image. Printer print pattern inspection device.