Hose printing quality detection technology develops with the development of people's requirements for printing. From the initial subjective visual evaluation to an objective evaluation based on a certain theory, from off-line detection to present online detection, its detection technology has been continuously innovating and developing.
First, the content and accuracy of hose printing quality inspection
1. Analysis of detection content
The main contents of the routine inspection of hose package printing are as follows:
(1) Use color blocks and lines to check the printed colors and registration;
(2) Judging the overall quality of the hose (including color, gradation, registration, surface condition, etc.). Mark the defective products and sort them in time. In addition, pay attention to the incorporation of other hose products of different specifications.
(3) Counting inspection, that is, checking whether the number of hoses is enough, usually one box per 500.
2. Requirements for testing the printing effect of hose packaging
(1) Body part:
Surface particles: The surface particle size must be greater than 0.03mm, but the particle color must be different from the tube body color to detect. There is no obvious difference between the particle color and the tube body.
Surface scratches: Scratches of the tube over varnish, dumb oil can be detected.
Color point: there are other color points on the solid color tube body, as shown in Figure 1. If the color points in the glow tube body detection cannot be detected, or the same color noise points cannot be detected, as shown in Figure 2.
Figure 1 Heterogeneous noise in pure colors
Figure 2 Same color noise
Figure 3 Defects such as carbonization point, foreign matter, etc. during high temperature die casting
(2) Printing (or silk screen or bronzing):
a), ghost / fluff / broken line / less dot / white dot / overprint;
b) the pattern (text) is clear;
c) Defects in bronzing. Gold loss means incomplete bronzing. The thickness is that the lines or text are embossed or not embossed.
d), plug (for screen printing process) / layout height / centering;
e), printing pattern shift, as shown in Figure 4;
Figure 4 Printing pattern shift
Testing accuracy and design parameters of hose printing quality
Table 1 Detection errors and design parameters
The accuracy of contrast between standard images and images of the printed products under inspection is a key issue for inspection equipment. Generally, inspection equipment collects images through a lens. In the middle part of the lens range, the image is very clear, but the image at the edge may be Ghost images, and the detection results of ghost components will directly affect the accuracy of the entire detection.
Third, functional requirements analysis of hose packaging printing quality inspection
The hose package printing quality inspection system must have the following functions because it needs to meet the functions of high-speed inspection and inspection sorting statistics:
1. Constant mechanical transmission and data acquisition function
This is to ensure the conditions of high-speed assembly line operations, to achieve continuous production; and to ensure that the image data is collected from time to time, providing the prerequisites for detection and analysis. Data collection is the basic condition to ensure the normal operation of the system, including two parts: the collection of system basic parameters and system equipment state parameters; the print image data collection.
2. Analysis and processing functions
Collection and processing can ensure the continuity of the system. The entire collection and processing analysis process is controlled within 0.1s. Statistical analysis of the quality of the product is provided to provide parameters for the improvement of the printing process.
3.Quick sorting and print sorting functions
The correct and timely sorting guarantees the continuous, stable and reliable system. The integration of the detection and sorting system into the entire printing production is a continuation of the entire function of the system; the sorting function of good and defective products is an extension of the system's function and is an extension of the entire production The process needs are supplemented.
4.Result statistical analysis function
Manufacturers need to know the quantity of qualified products, the yield rate, and the causes of defective products from time to time, and provide a basis for adjusting the parameters of the production process.
5. System equipment status detection
In order to ensure the normal operation of the system, it is necessary to understand the operating status of the equipment and the operating status of each subsystem at all times.
6.Standard template making function
This function can ensure the versatility of the system, as well as the detection of emerging new products.
7. System test and installation and debugging subsystem
Provide equipment debugging and installation debugging for the normal operation of the system, detect the installation attitude and other environments, and provide appropriate software and hardware foundations for the normal operation of the system.
Fourth, the implementation of software detection
As the soul of the system, the software is the most important component in the quality inspection system of hose packaging printing. Its main function is to realize the detection of the characteristics of specific targets through the processing, analysis and recognition of images. Here, matching tools, feature analysis tools, calibration tools, positioning measurement tools, character recognition tools, and color analysis tools are used to complete the image discrimination.
Figure 5 The acquired image
The main function of image processing and analysis tools is to enhance the collected images, which is convenient for subsequent professional visual tools to identify and understand. Histogram analysis, filtering operation, morphological operation, contour extraction, geometric transformation, and spatial color transformation are generally used to complete the analysis and processing of the image.
To determine whether the image meets the template requirements, more specialized technical tools are needed. Position calibration tools for position discrimination to achieve the conversion of image coordinates to spatial coordinates; area positioning tools and geometric positioning tools to identify other features; measurement tools and two-dimensional feature analysis tools; character recognition tools for character processing, Barcode or color analysis and identification tools; and reference tool performance comparison detection tools, etc., use these technical tools to complete the image processing and comparative analysis of hose packaging printing quality inspection.
Figure 6 Comparison of template and acquired image
Five, online detection
The machine vision system for the on-line inspection of the printing of hose packages is a high-performance integrated system based on the machine vision platform, which performs functions such as real-time detection, tracking, alarm, rejection, and information statistics. It is mainly composed of linear array vision sensor, mechanical transmission, rejection device, industrial control machine and single-chip micro-processor. It integrates high-speed automatic detection, measurement, statistics and processing. Maximum detection speed: 60Pcs / min, detection tube diameter: D16mm-D60mm, round tube length: L35mm-L205mm detection accuracy: 0.072mm, mainly used to track and identify stains, burrs, printing errors, Various printing quality defects such as printing offset, missed leakage, and bronzing offset; can greatly improve the finished product qualification rate and production efficiency of hose packaging printing, and ensure hose printing quality.
This system is a highly integrated optical-mechanical-electrical integration system with mechanical transmission, rejection device, optical lighting system, image acquisition system, constant processing analysis system, industrial control system, and automatic control system. The strong and weak current system realizes the function of unmanned automatic detection operation and automatic fault handling and alarm function.
1.Test case 1
Product: Mentholatum Multi-Activating Cleansing Milk 1000
Speed: 53 pcs / min
Location: Liying Plastic
Time: October 12, 2013
The hose inspection machine commissioned by Huatong Shengshi was commissioned by Tongcheng Lixing to be installed in August 2013, and completed the initial installation and commissioning work.
In order to further verify the stability of the machine in actual production, special tests are arranged, and the test objectives are:
Seek ways to reduce false alarm rates under current conditions;
Whether it is missed.
Figure 7 Complete image acquisition
Figure 8 Reset the template
a), blank position improves detection accuracy;
b), increase the number of template learning samples, 10;
c), learn the false detection content into the template at any time;
d) Reset QS and barcode position;
(2) Test results
Figure 9 Test interface and results
Table 2 Test results show
Conclusion: first test conclusion
1) The false detection rate meets the preliminary design requirements;
2), scratches and near-color defects can be detected well;
3), no missing inspection;
4), overprint, printing offset can be detected well;
5), glitches and chromatic aberration cannot be detected;
6), the detection speed is relatively slow and does not meet the requirements;
8) The brightness of the background is not uniform;
9), no positioning color code or color code is serious;
10) 、 Bronzing false positives.
Modifications after the first test:
1) Misalignment of misalignment-the vibration of the machine itself causes the instability of image acquisition, leading to batch misreports;
2) Printing misalignment-false positives due to printing misalignment;
Solution: Add edge control function, set deviation tightness through software, and prevent misreporting of deviation.
3) Background lightness-the lightness and darkness of the tube body are inconsistent when the camera is taking pictures, and large deviations may easily lead to false positives;
After adjusting the camera and light source many times and modifying the software, it was found that its lightness and darkness still existed, which was inexplicable, and accidentally discovered that the vibration of the die-casting machine on the same floor was very large, and its vibration caused the equipment to take images The intermediate process has shadows, which is the key to the different light and dark background colors;
Solution: cushion rubber cushion under the device and adjust the light source multiple times to achieve a moderate position and then fix;
4) False positives of hot stamping-because hot stamping products will have reflective dots on the sides of the image when taking images, it is easy to cause false positives and missed prints.
Solution: Set the parameters to relax the hot stamping characters through the software. For example, you can refer to the number of false positives.
5) Varnishes—burrs, strains, scratches, pits, bubbles.
Solution: By modifying the software calculation method.
6) .Pull body pulling, broken, internal organs;
Solution: Because the tube defects cannot be captured clearly, the system cannot judge, and the tube cannot be taken, so it cannot be detected. By repeatedly setting the learning template, it can finally detect the strain and abrasions.
After adjustment, the product was tested for the second time:
Test report after changing the strip light source to tunnel light:
After the strip light is replaced with tunnel light, the image acquisition effect is ideal, and the camera's white balance is re-corrected. The overall hue of the image is uniform, and there are no color vertical bars, but the lightness and darkness cannot be restored to the color of the tube itself. The current light source does not affect the detection accuracy, and there are no false positives caused by the lightness and darkness. The image in FIG. 10 is uneven in lightness and darkness. The vertical bars are also very obvious. The image in Figure 11 has even brightness and darkness, and the overall image is dark. The image has vertical bars.
Figure 10 Image acquisition under a bar light source
Figure 11 Image acquisition under tunnel light source
After the white balance correction, the image picture is very uniform, the brightness is improved, and there is no vertical bar when viewed in a single channel, but there will still be golden spots in the hot stamping bar. This golden spot needs to be adjusted by parameters to reduce false alarms, as shown in Figure 12.
New area detection-pull bar function. When drawing the detection area, set the single-channel pull bar detection in the area properties, and set the pull bar contrast bar length in the accuracy setting. Figure 13 Pull bar detection channel, Figure 14 Pull bar parameter settings.
Figure 12 Image collection of white balance adjustment under tunnel light source
Figure 13 Pull bar detection channel
Figure 14 Pull bar parameter settings
In the best state, re-do the drawing template and use the pull bar detection function. It was found that there were no false alarms caused by uneven background colors. The effect of detecting light black lines and black burrs was very good. FIG. 15 is a defect detection set.
Figure 15 Defect collection detection
Added the "area number" algorithm to the area detection. It is used in another hose product to mainly check the overprint misalignment between characters. In the accuracy setting, use the "measurement" parameter setting for directivity. Offset control, Fig. 16 is the area number window for area detection, and Fig. 17 is the defect parameter setting.
Figure 16 Area number window for area detection
Figure 17 Defect parameter settings
Mainly detect the misregistration of the text in groups 1 and 5. The text in group 1 is severely misaligned from left to right. This area can be set according to normal detection. As long as the text deviates from the detection area drawn, it will report missing or dirty spots , Can accurately detect misalignment, no false positives and no missed detections. The overprinting of the yellow English words in the 5 groups and the white bars on both sides is severely misaligned. The customer's request is that as long as the white bars are not adjacent to the yellow English words, they are considered positive.
There is a gap of 3mm between the two pieces of the standard good product. The actual deviation of the goods produced is inconsistent, and the severe deviation is even less than 0.5mm. Such cases are more common. Relying on the detection of the normal drawing area and relying on missing prints and dirty spots to control will cause a lot of false positives, and the relative rejection rate will be relatively increased. At present, the "number of areas" algorithm is used to detect the background of the white lines and yellow English characters before each cut is drawn (see Figure 18). The number of lines is used to control. When the lines and English letters are close to each other, The number will be reduced. In area detection, the number of areas is calculated and compared with the set number, it is easy to detect the difference.
Figure 18 Template mapping
FIG. 19 is an actual calculation detection map. The relative position of the algorithm in area calculation and the accuracy of edge feature calculation will be high. Therefore, in the subsequent detection, a large number of comparisons and area comparisons are used for detection.
Figure 19 Actual calculation detection diagram
Conclusion: The image taken under the tunnel light source has uniform pixels without distortion, and there are no black bars or vertical bars. The varnish scratches and the colored varnish burrs have obvious contrast and good effect. The gold dots in the hot stamping bars need to be more referenced to reduce false positives (362 tubes were tested in the same batch. Due to the number of 13 gold dots in the hot stamping bars, the false positive rate was 3.5%. The template learned 60, if you increase the number of learning will definitely reduce the false alarm rate). The pull bar function detects light black lines well. The overall judgment is that the lighting environment of the tunnel light is more stable and uniform than that of the strip light, and it is more suitable for detecting light-colored varnish products.
2.Test case 2
Product: Pink varnished oil tube, bronzing. As this product is tested for the first time, the varnish on the pink bottom has extremely high requirements for the light source. During the detection process, it is necessary to continuously modify and learn.
Table 3 Test content and data
Missing inspection analysis: This batch of tests is mainly for missing inspections. Most of the missed inspection types are varnish defects. On the one hand, varnish defects are not stable during imaging and cannot be detected 100%. Any varnish defects can be clearly identified when imaging Anything you see can be checked out. On the other hand, because the detection accuracy is relatively loose, when the contrast of the detection accuracy is greater than the actual contrast, it will be judged as a good product, resulting in missed detection. After resetting the accuracy, it can be detected.
False alarm analysis: 430 machine-tested waste products were spot-checked. There were 24 good products. The 24 good products were re-tested by the machine. Eighteen were false alarms caused by dust and color deviation on the tube during inspection. There are 6 actual dirty spots, and the area of the dirty spots is larger than the standard 0.03mm2, which is actually an acceptable defect in the production process.
Figure 20 Typical defect map
About the technical improvements needed for this test:
The installed brushes and blowing effect of the machine are very good, greatly reducing the dust on the pipe body and reducing the false alarm of dust by more than 90%. However, at present, the air outlet has only one air pipe and can only be blown to a certain point. No There may be dust in the place where it is blown. It is recommended to change to a whole row of blower ports to blow the entire pipe evenly.
The camera and the light source bracket are re-processed. The camera and the light source can be adjusted up and down. The closer the light source and the illuminated object are, the better the effect of image acquisition is, ensuring that the image color is uniform, so that the dirty line can use the pull bar function.
Some products cannot be triggered when the machine is tested due to severe color shift or products without color shifts. The system naturally cannot be detected and needs to be improved.
There is a conflict point between the accuracy level of using the dirty point control and the acceptable range. The accuracy of the control using the dirty point is relaxed to a larger range and level, and a more delicate precision control is achieved based on the detection.
The research on the printing error detection of hose packaging has gone through nearly three years before and after. It has been investigated and investigated in Beijing, Shanghai, Shenzhen, Xiamen and other places. Each base and its production workshop had a detailed understanding, and based on the needs analysis, the machine vision technology and the optical-mechatronics technology were designed. The methods and methods of using machine vision technology, optical-mechanical-electrical integration, and computer automation control integration technology to solve the problem of tube packaging printing quality inspection are proposed, especially a breakthrough in the field of tube printing online inspection.
The developed tube packaging printing error and leakage detection system perfectly solves a series of problems currently faced by manufacturers such as labor shortage, detection omission, low detection accuracy, and easy and slack labor. The inspection system built on the basis of Shenzhen Huatong Shengshi Technology Co., Ltd.'s die-cutting visual inspection and folding machine inspection has accumulated valuable experience for the visual inspection of hose packaging and other printing packaging in the future.