JPH0573584B2 - - Google Patents

Description

[Detailed description of the invention]

(Technical Field of the Invention) This invention automatically eliminates defects such as ink scattering, doctor streaks, and dust when patterns are periodically and continuously printed on packaging materials such as paper, aluminum, and plastic. The present invention relates to a method for detecting data even when printing is in progress. (Technical background of the invention and its problems) Generally, it is difficult to detect defects that appear in spots when printing patterns, etc. For this reason,
The equipment is stopped at the beginning or end of winding, and only a small portion is visually inspected. However, since it is not a 100% inspection process, defective products may sometimes be mixed in, which often causes problems. In order to solve this problem, recently a pair of patterns N1 and N2 (not necessarily adjacent patterns in this case) on a sheet N printed with the same continuous pattern as shown in FIG. Sensor head A and B
An apparatus has been proposed in which defects on a picture are determined by simultaneously performing photometry and photoelectric conversion, and then calculating the difference between the output voltages of a pair of photosensor heads A and B in a limit box 14. A concrete example of this will be explained with reference to FIG.
As shown in the figure, the reflected light from the patterns N1 and N2 is
After being input to optical sensor elements 16A and 16C through light receiving elements 15A and 15B such as optical fibers and converted into voltage, the gain is appropriately adjusted by amplifiers 16B and 16D, and the output voltage is input to differential amplifier 17. . Here, the light receiving element 15
Since A and 15B are located on the same patterns N1 and N2, if the patterns are normal, their reflected lights should be equal and the output of the differential amplifier 17 should be zero. However, it is physically impossible for them to be exactly equal, so we set the amount in advance within the allowable range, and here we set this value as the upper limit.
dmaz and the lower limit is set as dmin, and input to the comparator 18 together with the output of the differential amplifier 17. If ink is scattered on one of the patterns and this exceeds the set tolerance range, the output of the comparator 18 will be "NO" and, for example, will output a "1" signal to indicate an error. If it is within the range of the upper and lower limits, it will be a "GO" of a good product and a "0" signal will be output. In this way, a method for determining defects in printed matter, etc. is realized. However, the method described above requires that the optical sensing conditions of the pair of optical sensor heads remain constant. For this reason, various preparatory work is required when manufacturing the device, such as selecting optical sensor elements to match their characteristics, using a differential amplifier with good output characteristics, and deterioration of the optical sensor element due to aging and dirt on the sensor head during detection operation. In addition, it is necessary to manually perform correction for characteristic mismatch due to drift of the differential amplifier, etc. This is complicated and requires skill, and in order to reduce this, it is necessary to sacrifice detection accuracy and expand the dead band width, which is the allowable range of the upper limit dmax and lower limit dmin. It has been desired to develop a method for detecting defects in printed matter using a pair of optical sensor heads that automatically maintains constant optical detection conditions during operation. (Objective of the Invention) This invention was made in view of the above-mentioned circumstances, and it is a pair of optical detection conditions that automatically compensate and keep the optical detection conditions constant even if there is a discrepancy in the detection characteristics due to drift or changes over time. An object of the present invention is to provide a method for detecting defects in printed matter using an optical sensor head. (Summary of the Invention) The present invention relates to a method for detecting defects in printed matter. In the method for detecting defects in printed matter in which the same pattern is periodically and continuously printed, a pair of optical sensors are used. A pair of optical sensor heads arranged in groups according to the size of the picture are arranged perpendicularly to the moving direction of the printed matter, facing a pair of adjacent identical pictures, and A plurality of optical sensor heads are used simultaneously, and pulses are output from a pulse generator connected to a drive section of the plate cylinder, and each analog signal obtained from each pair of optical sensors in the pair of optical sensor heads is The output voltage difference of the analog signal is processed so as to correspond to the pulse corresponding to one rotation of one cylinder of the plate cylinder and its frequency division pulse every time the same picture is conveyed one cycle, and the output voltage of the analog signal is After digitizing the difference, one period of the digital signal of the pattern is stored, and the difference between the digital signal of the next pattern and the stored digital signal is calculated periodically. The pattern is continuously extracted in accordance with the picture to be conveyed, and the sign and ratio with the previously registered pattern are determined. If it is determined to be defective, the pulse or frequency division of the defective part is determined. Identifying the type of defect by counting the number of pulses,
The sign determination, the ratio determination, and the type of defect are displayed. (Embodiment of the Invention) FIG. 1 is a block diagram showing an example of a device for realizing the method of the present invention, which is entirely computerized, and when a predetermined key on a keyboard 1 is operated, information based on a start signal SS1 is displayed. At the same time as being displayed on the CRT 2, a signal SS2 is input to the main CPU 4. As a result, the same consecutive patterns N1, N
2... is printed at a predetermined pitch, and the driving unit 6, which includes a motor or the like, is driven, and winding of the sheet N on which the patterns N1, N2... are printed is started. When the winding is completed by a predetermined amount, the scanning head 5 detects a register mark that instructs the start of detection, and the register mark signal SS3 is outputted, and at the same time, a pulse generator 7 consisting of a pulse encoder etc. linked to the drive unit 6 outputs a signal for one cycle. A pulse corresponding to one rotation of one cylinder of the plate cylinder connected to the drive unit 6 corresponding to the picture and a pulse signal SS5 obtained by dividing the pulse are input to the main CPU 4 and further stored in the storage unit 3. That will happen. On the other hand, a pair of photosensors A1 and B1 in a pair of photosensor heads facing a pair of adjacent patterns N1 and N2
(There are up to optical sensors AN and BN depending on the size of the picture.) In order to detect whether the detection area in charge of the pair of pictures N1 and N2 is normal, the light sensor collects the reflected light from the picture and creates a voltage. It is converted into signals SS7 and SS8. Thereafter, the output voltage difference between the two is inputted to a differential amplifier 8 composed of an IC, etc., and is calculated and processed as an analog quantity, and the analog quantity of the difference is converted into a digital signal of a digital quantity corresponding to the analog quantity in the A/D converter 9.
After being converted to SS10 and input to the local CPU 11 for internal processing, the signal SS11 is sent to the storage unit 1.
Output to 0 and send processing signal SS to main CPU 4
12 will be input and will be collectively processed. In this case, the digital signal is stored for one period (one cylinder) of one picture, and the digital signal is always stored for one cycle of one picture.
By taking the difference between digital signals delayed by a period (cylinder), the difference is always doubled. Therefore, by recognizing and processing the temporal pattern of this difference in the main CPU 4, it is possible to obtain pattern detection information (for example, how much of a defect is the density and color of a certain area). Also, there are patterns N1 and N in the pair of photosensor heads A and B.
A pair of optical sensors A1, B1~
A large number of AN and BN are grouped together, and after the detection information for the patterns N1 and N2 of these printed materials is once processed by the local CPU 11, the main CPU 4
By collecting and processing the data collectively, it is possible to detect unforeseen circumstances, such as when one of a pair of optical sensor heads A and B fails during operation and cannot be detected, or when the output of a pair of optical sensor heads A and B fluctuates for some reason. However, by using this method, the system can always detect the situation, visualize the contents on CRT2, and monitor the situation one by one. In such a configuration, its operation is shown in Figure 2~
This will be explained with reference to FIG. 2A to 2C are flowcharts for explaining the operation, and FIG. 3 is a time chart showing a situation in which a pattern is detected according to this flowchart. Before operation, a sheet N on which the same patterns N1, N2, N3, etc. are printed periodically as shown in FIG. It is placed in a state where it is transported by cylinder. In this state, first, a predetermined key on the keyboard 1 is operated to send a start signal SS1, and the drive unit 6 is driven to wind up the sheet N, and the patterns N1, N2, N3, ... are sequentially 11 in the S direction. When the pattern moves for one period (one cylinder) and the register mark instructing the start of detection is identified, pulses are counted, and as shown in the figure, the pattern N1
Detection is started by a pair of optical sensor heads A and B positioned on
S1). In this case, the pair of optical sensor heads A and B are arranged facing each other on the adjacent patterns N1 and N2 as shown in the figure so that the detection area of the pair of optical sensor heads A and B can be set for one cycle of one pattern each. As a result, the output waveforms A(t) and B(t) of the analog signals from the optical sensors A1, B1 to AN, BN are also shifted in phase by one cycle of one picture. The method of this invention focuses on this point. By the way, in general, if the pattern is good like pattern N1, depending on the shading of the detection surface, for example, the third
A standard waveform such as A(t) between t0 and t1 in Figure B is output, but if there are spot-like defects a and b in some parts like pattern N2, the part a is brighter than the non-defective part.
Then, the dark portion b is output in a convex shape as shown in A(t) between Bt1 and t2 in the figure, and the dark portion b is output in a concave shape (see A, B, and C in FIG. 3). Hereinafter, an arbitrary i-th pattern Ni will be explained with reference to a mathematical formula. In this way, the i-th pattern Ni is placed on photosensor head A, and the pattern (Ni-1) is placed on photosensor head B.
is sent, and the reflected light is photoelectrically converted by a pair of photosensors Ai and Bi, resulting in output waveforms ai(t) and ti-(ti-1).
bi(t) is obtained (steps S1 and S3, FIG. 3 B and C), and then the difference between the output waveforms bi(t) - ai(t) is obtained by the arithmetic processing of the differential amplifier 8 ( Step 4, Figure 3D). Next, one picture is divided into N parts (the size of one pulse is not a unit of time, but a unit of length), and sampling is performed for each divided section (step S5). The difference between the output waveforms is digitized, and the difference in digital amount C=Cij (Pj) is obtained as the j division section of the n divisions of the i-th picture Ni (step S7, FIG. 3E). This is temporarily written and stored in the RAM of the storage unit 10 for use in the next calculation, and this is set to D=-1 (Pj) (hereinafter, when it is to be played back after being stored, is added). Steps S8, S9, Figure 3 F), then 1 of 1 picture
When the sheet N is conveyed for one cycle (for one cylinder) and the pattern Ni detected earlier by the optical sensor head A is placed on the optical sensor head B, the difference in digital amount C = Cij (Pj) is calculated in the same way as above. , locally store D stored in the storage unit 10 and this C.
Processing is performed within the CPU 11, and the difference C-D=dij(Pj)=
Cij(Pj)-(-1)(Pj) is obtained (step S11, FIG. 3G). Next, set the upper limit of the allowable range for whether the pattern is normal or not by using the insensitive upper limit dmax.
E: eij (Pj) = |dij (Pj) | −dmax≧
If it is 0, it continues as a processing target, and if ≦0, it is not considered and returns to step S5, and -1 is set.
dij (steps S12, S13, S14). after that,
Whether the results of performing the same detection on the pattern N1 with a pair of sensor heads A and B are equal or not, respectively.
The pulse number is determined by comparing the number of pulses. In other words, if Pi=, processing continues, Pj≠
If Pj, it is assumed that a positional shift in detection has occurred and is carried over (step S15). In this way, in order to check whether the pattern Ni is a good product or how defective it is, sign determination and ratio determination are performed. In other words, if it is a defective product, the positive and negative signs are always output with the wrong sign, so the pattern judgment shown in the table below is performed (step
S16, S17).

[Table] Here, patterns corresponding to various defects are registered in advance in the internal storage section of the main CPU 4, and it is determined which pattern corresponds to the defect (step S18). This makes it possible to identify what kind of defects the pattern Ni has. If not applicable, prepare to update the memory of the carried forward amount (step
S19). If applicable (step S20), parts unnecessary for the determination, such as - in the sign determination in Table 1, are
-a at the end of the output pattern of a, 2a, -a is unnecessary for judgment, so in order to eliminate erroneous processing, it is deleted and the RAM of the storage unit 10 is cleared, and the memory is updated for the next pattern (Ni + 1). Prepare (step S21). Next, in order to check whether the entire defect detection system is operating normally, after detecting a certain number of patterns N1 to Ni, the number of defects is detected. That is, the total amount Σ|eij| of the aforementioned |dij| is compared with the total upper limit value e that has been registered and set in the internal storage section of the main CPU 4 in advance (step S23).
If Σ｜ei｜−e≧0, it is an abnormal occurrence of a defect,
It is determined that this is not a printing error but is caused by a misalignment or meandering of the sheet N, and the machine is re-inspected (step S24). Also Σ｜ei｜−e＜
If it is 0, it is considered a normal printing error and the main
This information is transmitted to the CPU 4 and displayed on the CRT 2 (step S25). At the same time, it is determined whether the defects in the patterns N1, N2, etc. are within the acceptable range for the product, and whether they are major, medium, or minor defects; in other words, whether the defects are continuous or one-off. In order to
Defects are checked for pulses...Pj-1, Pj, Pj+1, etc. in adjacent sections centering around the divided j-divided section (step S26), and if there is a defect, the number of adjacent pulses is counted (step S26).
S27), if only Pj is counted due to spot defects, such as dust, etc. (step
S29), or is it a case where Pj-1, Pj, Pj+1 are counted due to a narrow-area defect, such as a doctor's line (step S30), or a wide-area defect,
For example, ink stains are spread over a wide area...Pj−
1, Pj, Pj+1, etc. are connected and counted in large numbers, etc. It is determined to which type the defect belongs, and the number of occurrences of defects is counted in one cycle of one pattern (steps 32, S33, S34). ), the defect location and type are displayed on the CRT 2 to notify defect information of the picture (step S35). Furthermore, in order to detect the next pattern (Ni+1), a signal is sent to the main CPU 4 for general processing (step S36), and the drive unit 6 rotates by one cylinder. By the way, in parallel with this, the process after the above-mentioned memory update preparation (step S21) is performed continuously for one cycle of one picture, but if an abnormality occurs during the process and the process cannot be performed, the process returns. Then, the process returns to sampling division (step S5), and when the processing is completed, the carried over/erased data of the pattern Ni is cleared from the RAM of the storage unit 3, and the first register mark is identified in order to detect the next pattern (Ni + 1). Return to (step S1) and
The same treatment as for Ni will be performed. In this way, all the patterns N1, N2, N3...,
By repeatedly processing steps S1 to S38 in the flowchart of FIGS. 2A to C for each NN, a pair of photosensor heads A and B sequentially memorize one period of one picture, and compare and update the images. By processing, defect information of patterns N1, N2,...NN printed continuously on sheet N is transferred to CRT2.
etc. will be displayed. Now, in the case of the above-mentioned detection, the detection output of the pair of optical sensor heads A and B does not change, but the drift of the differential amplifier 8 or the aging of the optical sensors A1, B1 to AN, BN may occur during operation of the device. Even if there is a variation in the detection characteristics due to changes in the detection characteristics, by using the above method, the variation can be automatically corrected and an accurate detection output can be obtained. That is, FIG. 4 is a time chart for explaining the above situation.
As shown in the figure, if optical sensor heads A and B are raised by f and lowered by g relative to the reference amount, the deviations f and g are normally accumulated and magnified as errors by performing calculations one after another. Conventional methods caused overestimation. However, according to the method of this invention, FIG.
By performing the calculation process of C-D in step S11 according to the flowchart of
drift or optical sensors A1, B1 to AN,
The deviations f and g due to aging of the BN are canceled and automatically corrected, and only the defects in the pattern (a, b, c, etc. in Fig. 4G) are extracted. That is, Optical sensor head A: Output waveform A(t)
...(Step S2) Optical sensor head B: Output waveform B(t)
...(Step S3) C=B-A:B(t)-A(t) ...(Step S4) D:(t)-(t)...(Step S10) C-D: {B(t)-A (t)}−{ B (t)− A (t)}
...(Step S11), and here the deviations f and g occur, A(t)→A(t)+f, B(t)→B(t)-g, and by step S4, C-D=[{B (t)−g}−{A(t)+f}] −[{ B (t)−g}−{ A (t)+f}] = {B(t)−A(t)}−{ B ( t)− A (t)}
(Step S11) Therefore, even if the detection characteristics vary due to the deviations f and g caused by any factors, they are canceled out in the calculation process, and only the defects a, b, c, etc. of the desired pattern are found. To explain this with reference to the time chart in Figure 4, for example, between t1 and t2, the output waveform fluctuates overall by A(t)+f and B(t)-g by the deviations f and g, respectively. Similarly, the picture defects a and b are moved by -fg in the calculation process B(t)-A(t) (step S4) to become -a-f-g and b-f-g ( Figure 4D). As shown in the figure, after A/D conversion, it is digitized (Fig. 4E), and further arithmetic processing C-D (step
In step S11), only the defect information of the same pattern as shown in FIG. 4G, which is the same as that shown in FIG. 3G without the above-mentioned variation, is obtained. Therefore, during operation of the device, the drift of the differential amplifier 8 or the optical sensors A1, B1...
Even if there is a change in the detection characteristics due to changes in AN and BN over time, the method of the present invention automatically corrects it through internal processing to ensure accurate detection. Furthermore, if one of the pair of photosensors A and B fails and becomes unusable, for example, if photosensor head B fails and the output waveform B(t) cannot be obtained (Fig. 5C), the same According to the method of the present invention, which detects the pattern Ni and compares the images at different times, since the output waveform A(t) is obtained from the optical sensor head A, defects in the pattern can be detected. FIG. 5 is a time chart for explaining this, and shows calculation processing B(t)-A(t) (step S4).
becomes the output waveform of the optical sensor head A (Fig. 5D), is digitized after A/D conversion (Fig. 5E), and is converted into the 5th waveform by arithmetic processing C-D (step S11).
As shown in Figure G, only the picture defects a, b, and c are extracted, but in this case, unlike the two cases described above, the pattern determination (steps S16 and S17) is as shown in the table below (Table 2).

[Table] The above Table 2 is different from the pattern judgment table (Table 1) when using a pair of photosensor heads A and B described above, but this differs from the pattern judgment table (Table 1) for sign judgment and ratio judgment depending on the type of pattern defect. This is because the standards are different, and the main
By replacing the internal memory of the CPU 4, it is possible to easily and accurately detect defects in patterns, and this information can be processed into images using a CRT 2 or the like. In this way, by using a pair of photosensor heads A and B in each of the three cases above, and adjusting the gain appropriately with a differential amplifier using the output difference as an analog quantity, the defect in the pattern can be enlarged to the desired amount. Alternatively, it is also possible to reduce the size. Furthermore, even if noise components due to fluctuations or the like are mixed in during arithmetic processing, only pattern defect information can be obtained in the end, resulting in a highly accurate printed matter defect detection device. (Effects of the Invention) As described above, according to the method of the present invention, in a printed matter defect detection device for detecting defects in printed matter in which the same pattern is periodically and continuously printed, the quality of the performance of the device is determined. It has many advantages, such as being able to automatically detect and visually display the types of defects in printed matter with high precision without requiring maintenance during operation.

[Brief explanation of the drawing]

FIG. 1 is a block diagram showing an example of an apparatus for implementing the method of the present invention, FIG. 2 A, B, and C are flowcharts for explaining the method of the present invention, and FIG.
Figures 4 and 5 are time charts for explaining the method of the present invention, and Figures 6 and 7 are for detecting defects in printed matter when the same pattern is printed continuously. FIG. 2 is a diagram for explaining a detection device. 1...Keyboard, 2...CRT, 3...Storage unit, 4...Main CPU, 5...Scanning head, 6...Drive unit, 7...Pulse generator, A
1, A2, ~AN and B1, B2, ~BN... optical sensor, 8... differential amplifier, 9... A/D converter, 10... storage section, 11... local CPU,
N1, N2, ~NN...Picture, N...Sheet with the picture printed on it, A and B...Photo sensor head, 1
4... Control box, 15A and 15B... Light receiving element, 16A and 16C... Light sensor element, 16
B and 16D...amplifier, 17...differential amplifier,
18...Comparator.

Claims (1)

[Claims] 1. In a method for detecting defects in printed matter in which the same pattern is periodically and continuously printed, a pair of optical sensors are arranged in groups according to the size of the pattern. 1 A pair of optical sensor heads are disposed perpendicularly to the moving direction of the printed matter, facing a pair of adjacent identical patterns, and the pair of optical sensor heads are used simultaneously, and the driving unit of the plate cylinder is connected to the optical sensor head. Pulses are output from the connected pulse generators, and the output voltage difference of each analog signal obtained from each pair of optical sensors in the pair of optical sensor heads is calculated every time the same picture is conveyed one cycle. After processing the pulse corresponding to one rotation of one cylinder of the plate cylinder and its frequency divided pulse, and digitizing the output voltage difference of the analog signal, one digital signal of the picture is converted into one digital signal. Only the period is stored, the difference between the digital signal for one period of the next pattern and the stored digital signal is determined, and this difference is continuously patterned and extracted in correspondence with the pattern that is periodically conveyed. Then, it performs sign and ratio judgments with the pre-registered pattern, and if it is determined to be a defect, the type of defect is identified by counting the number of the pulses or frequency-divided pulses at the defective location. With,
A method for detecting defects in printed matter, characterized in that the code determination, the ratio determination, and the type of defect are displayed.