Machine vision is the use of machines instead of the human eye to make measurements and judgments. Machine vision system refers to the machine vision products (i.e. image acquisition device, divided into CMOS and CCD two kinds) will be taken into the target into the image signal, transmitted to the special image processing system, according to the pixel distribution and brightness, color and other information, into digital signals; image system of these signals to carry out a variety of operations to extract the characteristics of the target, and then according to the results of the discrimination to control the equipment on the scene.
A typical industrial machine vision application system consists of the following parts: light source, lens, CCD camera, image processing unit (or image capture card), image processing software, monitor, communication/input/output unit. First of all using the camera to obtain the image signal of the target to be measured, and then through the A/D conversion into a digital signal transmitted to the special image processing system, according to the pixel distribution, brightness and color information, and other operations to extract the characteristics of the target, and then according to the preset criteria for judging the output results to control the drive actuators for the corresponding processing. Machine vision is a comprehensive technology, which includes digital image processing technology, mechanical engineering technology, control technology, light source illumination technology, optical imaging technology, sensor technology, analog and digital video technology, computer hardware and software technology, human-computer interface technology and so on. Machine vision emphasizes the practicality, requires to be able to adapt to the harsh environment of industrial site, to have a reasonable cost-effective, universal industrial interface, high fault tolerance and security, and has a strong generality and portability. It emphasizes more on real-time, and requires high speed and high precision.
The output of the vision system is not an image video signal, but an inspection result after arithmetic processing, such as dimensional data. The upper computer, such as PC and PLC, obtains the detection results in real time, and then commands the motion system or I/O system to perform the corresponding control actions, such as positioning and sorting. From the classification of the vision system's operating environment, it can be categorized into PC-BASED systems and PLC-BASED systems. PC-based systems utilize their openness, high programming flexibility and good Windows interface, while the overall system cost is low. Taking the American DATA TRANSLATION company as an example, the system contains a high-performance image capture card, which can generally be connected to multiple lenses. In terms of supporting software, there are several levels from low to high, such as the DLL for C/C++ programming under Windows95/98/NT environment, the visualization control activeX provides a graphical programming environment under VB and VC++, even the The object-oriented machine vision configuration software for Windows allows users to quickly develop complex and advanced applications. In PLC-based systems, the role of vision is more like an intelligent sensor, the image processing unit is independent of the system, and exchanges data with the PLC through serial buses and I/Os. The system hardware generally utilizes a high-speed dedicated ASIC or embedded computer for image processing, and the system software is solidified in the image processor, which is configured by a simple device similar to a gaming keyboard for the menus displayed in the monitor, or by developing software on a PC and then downloading it. PLC-based system reflects the high reliability, integration, miniaturization, high-speed, low-cost features, on behalf of the manufacturers of Japan's Panasonic, Germany Siemens and so on.
Germany Siemens has more than 20 years of experience in industrial image processing, SIMATIC VIDEOMAT is the first high-performance monochrome and color image processing system, and has become SIMATIC automation system is extremely important products. And SIMATIC VS710 launched in 99 years is the industry's first intelligent, integrated, with PROFIBUS interface, distributed gray-scale industrial vision system, which will be the image processor, CCD, I/O integrated in a small chassis, providing PROFIBUS networking (communication rate of up to 12Mbps) or integrated I/O and RS232 interface. What's more, configured via the Pro Vision parameterized software under PC WINDOWS, the VS 710 combines for the first time the flexibility of a PC with the reliability of a PLC, distributed network technology, and an integrated design, allowing Siemens to find the perfect balance between PC and PLC systems. Machine Vision System in Printing and Packaging The automatic print quality inspection equipment used in the detection system is mostly the first use of high-definition, high-speed camera lens to shoot the standard image, on the basis of which a certain standard is set; and then shoot the image to be detected, and then compare the two. ccd linear sensors will be converted to electronic signals for each pixel of the change in the amount of light, as long as it is found to be different from the standard image, the system will be compared with the standard image. The CCD linear sensor converts each pixel into an electronic signal, and whenever a difference is found between the inspected image and the standard image, the system considers the inspected image to be a defective product. Various errors in the printing process are only the difference between the standard image and the inspected image to the computer, such as smudges, ink spots, color differences, and other defects are included.
The earliest used for print quality inspection is the standard image and the detected image for gray-scale comparison of technology, the more advanced technology is based on RGB three primary colors for comparison. Fully automated machine inspection compared to human eye inspection, where is the difference? Take the human eye as an example, when we concentrate on a print, if the print contrast color is relatively strong, the human eye can be found, the smallest defects, is the contrast color is obvious, not less than 0.3mm defects; but relying on the human ability is difficult to maintain a sustained, stable visual effect. But another situation, if it is in the same color printing to find defects, especially in a light color system to find quality defects, the human eye can find the defects need to have at least 20 gray level difference. An automated machine can easily find defects as small as 0.10mm, even if they are only one gray level different from the standard image.
But in practical terms, even the same full-color contrast system has different abilities to identify color differences. Some systems are able to detect defects with large variations in contour sections and color shifts, while others can identify very small defects. For white cardboard and some simple style prints, such as Japan's KENT cigarette labels, the United States, Marlboro cigarette labels, simple detection may be sufficient, while most of the domestic prints, especially a variety of labels, with many features, with too many glittering elements, such as gold and silver cardboard, hot stamping, embossing, or varnishing prints, which requires that the quality inspection equipment must be equipped with enough to find very small Gray scale difference in the ability to find, perhaps 5 gray scale difference, perhaps more stringent 1 gray scale difference. This is critical for the domestic labeling market.
Standard image and the inspection of the printed image of the contrast is accurate is the key issue of the detection equipment, usually, the detection equipment is through the lens to capture the image, in the middle part of the lens range, the image is very clear, but the edge part of the image may produce a shadow, and the shadow part of the detection results will have a direct impact on the accuracy of the entire test. From this point of view, if only the full-width area of the comparison is not suitable for some fine prints. If the resulting image can be subdivided again, for example into 1024dpi X 4096dpi or 2048dpi X 4096dpi, the detection accuracy will be greatly improved, and at the same time, because of the avoidance of the edge of the part of the shadow, so that the detection of the results of more stable.
The use of inspection equipment for quality inspection provides real-time reports of the entire inspection process and detailed, complete analysis reports. Field operators can rely on fully automated testing equipment, timely alarms, according to real-time analysis report, timely adjustment of problems in the work, perhaps reducing the scrap rate will not only be a percentage point, the manager can be based on the analysis of the test results of the report on the production process tracking, more conducive to the management of production technology. Because of customer requirements, high-quality testing equipment, not only to stay in the check out the good and bad print, but also requires the ability to analyze the aftermath. Some of the quality testing equipment can do not only improve the qualified rate of the finished product, but also assist manufacturers to improve the process, the establishment of a quality management system, to achieve a long-term stable quality standards.
Gravure printing press position control and product inspection
A video image of the printed product is continuously captured by a camera set up on the production line at an adjustable speed of 30 frames/s or less. The images captured by the camera are first quantized, the analog signal is converted to a digital signal, and a key frame that effectively represents the content of the footage is extracted from it and displayed on a monitor. For a frame, the analysis of the still image can be used to deal with, through the size measurement and multi-spectral analysis can be identified on the video image of each color scale, to get the color scale spacing and color parameters of the color scale, as well as some other related.
Due to a variety of factors, there will be a variety of noise, such as Gaussian noise, pepper noise and random noise. Noise to the image processing brings a lot of difficulties, it has a direct impact on the image segmentation, feature extraction, image recognition, so the real-time acquisition of images need to be filtered. Image filtering requires the ability to remove noise outside the image, while maintaining the details of the image. When the noise is Gaussian noise, the most commonly used is the linear filter, easy to analyze and implement; but the linear filter is very poor for the filtering effect of salt and pepper noise, the traditional median filter can reduce the salt and pepper noise in the image, but the effect is not ideal, that is, the noise is sufficiently dispersed to be removed, while the noise close to each other will be retained, so when the salt and pepper noise is more serious, it is filtering the effect of the filtering is obviously worse. This system improved median filtering method. This method first obtains the median value of the noisy image window after removing the pixels with the largest and smallest gray values, then calculates the difference between this median value and the corresponding pixel's gray value, and then compares it with the threshold value to determine whether to replace the pixel's gray value with the obtained value.
Image segmentation detects each color scale and separates it from the background in this stage, and the edges of an object are reflected by gray-scale discontinuities L Edge types can be divided into two types, step edges, which have pixels on either side of them with significantly different gray-scale values, and roof edges, which are located at the turning point of the change in the gray-scale value from an increase to a decrease L In the case of step edges, the second-order directional derivatives are zero-crossed at the edges, and thus can be differentiated by differential arithmetic. zero-crossing, and thus the differential operator can be used as the edge detection operator. Differential operator type edge detection method is similar to the high spatial domain of high-pass filtering, there is an increase in the role of high-frequency components, this type of operator is quite sensitive to noise, for step edges, usually available operators are gradient operator Sobel operator and Kirsh operator. For roof-like edges the Laplace transform and the Kirsh operator are available. Since the color scale is rectangular and the difference in gray level between adjacent edges is large, edge detection is used to segment the image. Here, Sobert's edge is used for edge detection, which uses local difference operator to find the edge, and it can separate the color scale better. In the actual detection process, the color image edge detection method is used to select the appropriate color base (e.g., intensity, chroma, saturation, etc.) for detection. According to the characteristics of the type of printing press, i.e., the color of each color of the printing press and the characteristics of the plate, multi-threshold processing is carried out to obtain the binary map of each color.
The segmented image is measured, and the object is recognized by the measured value. Since the color scale is a rectangle of regular shape, the following features can be extracted: (1) the rectangle area is calculated from the pixels, (2) the rectangle degree, (3) the chromaticity (H ) and saturation (S ), and then the spacing between the color scales is obtained according to the number of pixel dots separating each color scale, which is compared with the set value to obtain the two Then the distance between the color markers is obtained according to the number of pixels between the color markers and compared with the set value to obtain the difference between the two. To adjust the relative position of the color rollers, thereby eliminating or reducing printing misalignment. In the feature extraction, multi-spectral image analysis of the image can quantitatively represent the color scale, such as the color of the pixels in the color number image, using the HIS format to get the color information of each color scale of the two parameters: chroma and saturation, in order to detect the quality of the ink. For each color binary value map and then statistical calculation or with the standard graphic sample matching, measurement of the printing process, such as ink chips and other parameters.
Printing machine from the uncoiler unwind running in turn through the printing unit for each color of the printing and drying, by the winding machine for the winding L Each color printing will be printed on the edges of the printed material for color printing color markers, the color marking line horizontal 10 mm, 1 mm wide, each adjacent color marking line should be parallel to each other in overprinting precision, vertical (longitudinal) giant 20 mm, set up by the camera on the production line The camera set up in the production line to continuously capture the video image of the printed products, through size measurement and multi-spectral analysis can identify the video image of the color markers, get the color markers spacing and color markers of the color parameter L If the two adjacent color markers spacing is greater than or less than 20 mm, then it is shown that overprinting deviation. The deviation signal is sent to the servo frequency conversion drive unit to drive the AC servo motor, so that the corresponding overprint correction roller ML up and down to extend or shorten the print material from the last unit of the printing plate roller to the unit of the printing plate roller to the stroke of the dynamic correction. In the modern packaging industry automated production, involves a variety of inspection, measurement, such as beverage bottle cap printing quality inspection, product packaging on the bar code and character recognition. This type of application is characterized by continuous mass production and very high demands on appearance quality. Usually this kind of work with a high degree of repetitiveness and intelligence can only rely on manual inspection to complete, we often see hundreds or even more than a thousand inspection workers in some factories behind the modern assembly line to perform this process, in the factories to increase the huge labor costs and management costs at the same time, still can not guarantee that 100% of the inspection pass rate (i.e., zero defects), and today's competition between enterprises, has not allowed even 0.1%. The competition among enterprises today does not allow even 0.1% defects to exist. Some times, such as the precise and rapid measurement of small dimensions, shape matching, color identification, etc., with the human eye can not be continuous and stable, other physical sensors are also difficult to use. At this time, people began to consider the rapidity, reliability, and repeatability of the results of the computer, thus introducing the robot vision technology.
Generally speaking, first of all, the CCD camera is used to convert the target to be taken into an image signal, which is transmitted to the special image processing system, according to the pixel distribution and brightness, color and other information, such as: area, length, quantity, position, etc.; finally, the results are outputted according to the preset tolerances and other conditions, such as: size, angle, offset, number, qualified / unqualified, yes / no, and so on. Machine vision is characterized by automation, objectivity, non-contact and high accuracy. Compared with image processing systems in the general sense, machine vision emphasizes accuracy and speed, as well as reliability in industrial field environments. Machine vision is extremely suitable for measurement, inspection and identification in mass production process, such as: identification of printed characters on the surface of IC, identification of production date on the top face of food packages, and inspection of label placement position. In the machine vision system; the key technologies are light source illumination technology, optical lens, camera, image acquisition card, image processing card and fast and accurate actuator and other aspects. In the machine vision application system; good light source and lighting program is often the key to the success or failure of the entire system; plays a very important role; it is not simply illuminate the object only. Light source and lighting program should be as much as possible to highlight the amount of object features; in the part of the object to be detected and those who are not important parts of the object should be as much as possible to produce a clear difference between the contrast; increase contrast; at the same time it should also ensure that the overall brightness is sufficient; changes in the position of the object should not affect the quality of the imaging. Transmitted light and reflected light are generally used in machine vision applications. For the case of reflected light should fully consider the relative position of the light source and optical lens, the texture of the object surface; the geometry of the object, the background and other elements. The choice of light source must be consistent with the desired geometry, illumination brightness, uniformity, spectral characteristics of light emission, etc.; but also consider the luminous efficiency and service life of the light source. Optical lens is equivalent to the lens of the human eye; it is very important in the machine vision system. A lens of the imaging quality is good or bad; that is, its aberration correction is good or not; can be measured by the size of the aberration; common aberration has a spherical aberration, comet aberration, like scattering, field curvature, aberration, chromatic aberration and so on the six kinds of aberration.
Camera and image acquisition card **** with the completion of the material image acquisition and digitization. High-quality image information is the original basis for the system's correct judgment and decision-making; is another key to the success of the entire system. In the machine vision system; CCD camera with its compact size, reliable performance, clarity and other advantages have been widely used. CCD camera according to its use of CCD devices can be divided into two categories of line array and surface array. Line array CCD camera can only get a line of image information; the object being photographed must be in the form of a straight line from the front of the camera to move through; to get a complete image; therefore, it is very suitable for a certain speed of uniform motion of the material flow of the image detection; while the surface array CCD camera can be obtained at a time the whole image of the information. Image signal processing is the core of the machine vision system; it is equivalent to the human brain. How to process and operate on the image; that is, the algorithm is embodied here; is the focus of the development of the machine vision system and where the difficulty lies. With the rapid development of computer technology, microelectronics technology and large-scale integrated circuit technology; in order to improve the system's real-time; a lot of work on image processing can be accomplished with the help of hardware; such as DSP, special image signal processing card, etc.; the software is mainly to complete the algorithms in a very complex, less mature, and need to continue to explore and change the part.
From the product itself, machine vision will tend to rely more and more on PC technology, and data acquisition and other control and measurement will be more closely integrated. And embedded-based products will gradually replace board-based products, which is a growing trend. The main reason is that with the rapid development of computer technology and microelectronics technology, embedded system applications are becoming more and more extensive, especially with its low-power technology has been emphasized. In addition, the vast majority of embedded operating systems are based on the C language, so the use of C high-level language for embedded systems development is a basic work with the use of high-level language advantages are to improve efficiency, shorten the development cycle, and more importantly, the development of products with high reliability, maintainability, easy to continuously improve and upgrade, and so on. Therefore, embedded products will replace the board-type products.
As machine vision is part of automation, without automation there will be no machine vision, machine vision hardware and software products are gradually becoming a collaborative production and manufacturing process at different stages of the core system, both the user and the hardware supplier will be the machine vision products as a production line information collection tool, which requires a large number of machine vision products to adopt standardized technology, intuitively said to be With the opening of automation and gradually open, can be based on the user's needs for secondary development. Today, automation companies are advocating the integration of hardware and software solutions, machine vision vendors in the next 5-6 years should also be not simply a supplier of products only, but gradually to the integration of solutions to the system integrator.
In the next few years, with the development of China's processing and manufacturing industries, the demand for machine vision is also gradually increasing; with the increase in machine vision products, technology improvement, the application of machine vision in China will be from the initial low-end to high-end. Due to the intervention of machine vision, automation will develop in the direction of smarter and faster.