Application Status of Machine Vision Technology

Machine vision is to measure and judge with machines instead of human eyes. Machine vision system refers to the machine vision products (that is, camera devices, which are divided into CMOS and CCD) that convert the photographed objects into image signals, transmit them to a special image processing system, and convert them into digital signals according to pixel distribution, brightness, color and other information. The image system performs various operations on these signals, extracts the characteristics of the target, and then controls the action of the field equipment according to the discrimination result.

A typical industrial machine vision application system includes the following parts: light source, lens, CCD camera, image processing unit (or image acquisition card), image processing software, monitor, communication/input/output unit, etc. Firstly, the image signal of the measured object is obtained by camera, and then it is converted into digital signal through A/ D conversion and transmitted to a special image processing system. According to the information of pixel distribution, brightness and color, various operations are carried out to extract the characteristics of the object, and then the judgment results are output according to the preset judgment criteria to control and drive the actuator to carry out corresponding processing. Machine vision is a comprehensive technology, including digital image processing technology, mechanical engineering technology, control technology, light source lighting technology, optical imaging technology, sensor technology, analog and digital video technology, computer software and hardware technology, man-machine interface technology and so on. Machine vision emphasizes practicality, requires adaptability to harsh industrial environment, reasonable cost performance, universal industrial interface, high fault tolerance and safety, and strong versatility and portability. It emphasizes real-time performance and requires high speed and accuracy.

The output of the vision system is not an image and video signal, but a detection result after operation, such as size data. After the upper computer, such as PC and PLC, obtains the test results in real time, it commands the motion system or I/O system to perform corresponding control actions, such as positioning and sorting. According to the operating environment of vision system, it can be divided into PC-based system and PLC-based system. The system based on PC makes use of its openness, high programming flexibility and good Windows interface, and the overall cost of the system is low. Take the American data translation company as an example, the system contains a high-performance image acquisition card, which can generally connect multiple lenses. In terms of supporting software, there are several levels from low to high, such as DLL programmed by C/C++ in Windows95/98/NT environment, activeX of visual control, graphical programming environment in VB and VC++, and even object-oriented machine vision configuration software in Windows, which users can use to quickly develop complex advanced applications. In the system based on PLC, vision is more like an intelligent sensor. The image processing unit is independent of the system, and exchanges data with PLC through serial bus and I/O. The system hardware generally adopts high-speed ASIC or embedded computer for image processing, and the system software is solidified in the image processor, and the menu displayed in the monitor is configured through a simple device similar to a game keyboard, or the software is developed on a PC and downloaded. The system based on PLC embodies the characteristics of high reliability, integration, miniaturization, high speed and low cost. The representative manufacturers are Panasonic in Japan and Siemens in Germany.

Siemens Germany has accumulated more than 20 years of experience in industrial image processing. SIMATIC VIDEOMAT is the first high-performance monochrome and color image processing system, and it has become a very important product in SIMATIC automation system. SIMATIC VS7 10 was introduced in 1999, which is the first intelligent, integrated and distributed gray industrial vision system with PROFIBUS interface. It integrates the image processor, CCD and I/O in a small box, and provides the networking mode of PROFIBUS (communication rate reaches 12Mbps) or integrates I/O and RS232 interfaces. More importantly, through the configuration of Pro Vision parametric software under PC WINDOWS, VS 7 10 combines the flexibility of PC, the reliability of PLC, distributed network technology and integrated design for the first time, which makes Siemens find a perfect balance between PC and PLC system. Application of machine vision system in printing and packaging Most of the inspection systems used in automatic printing quality inspection equipment shoot standard images with high-definition high-speed camera lenses first, and then set certain standards. Then take the detected image and compare the two. CCD linear sensor converts the light quantity change of each pixel into an electrical signal. After comparison, as long as the detected image is found to be different from the standard image, the system considers the detected image unqualified. All kinds of errors in the printing process are only the differences between the standard image and the detected image for the computer, such as stains, ink dot color difference and other defects are included.

The earliest technology used to detect the quality of printed matter is to compare the standard image with the detected image in gray scale, and the more advanced technology is based on RGB three primary colors. What's the difference between automatic machine inspection and human eye inspection? Take human vision as an example. When we focus on a printed matter, if the contrast color of the printed matter is strong, the smallest defect that the human eye can find is that the contrast color is obvious and not less than 0.3mm, but it is difficult to maintain a continuous and stable visual effect by relying on human ability. On the other hand, if we look for defects in printed matter of the same color system, especially in light color system, the defects that can be found by human eyes need at least 20 gray levels. Automata can easily find defects with the size of 0. 10mm, even if the defect is only one gray level away from the standard image.

However, in practice, even the same panchromatic contrast system has different color difference resolution. Some systems can find defects with large changes in contour and color difference, while others can identify tiny defects. For white cardboard and some simple printed matter, such as Kent cigarette labels in Japan and Marlboro cigarette labels in the United States, simple detection may be enough, while most printed matter in China, especially various labels, have many characteristics, such as gold and silver cardboard, bronzing, embossing or polishing printed matter, which requires quality inspection equipment to have enough ability to find tiny gray difference, perhaps five gray difference. This is very important for the domestic label market.

The accurate contrast between the standard image and the detected printed image is the key problem of the detection equipment. Usually, the detection device collects images through the lens. In the middle part of the lens range, the image is very clear, but the image in the edge part may produce ghost, and the detection result of the ghost part will directly affect the accuracy of the whole detection. From this point of view, if it is only the comparison of the whole region, it is not suitable for some fine printed matter. If the obtained image can be subdivided again, such as 1024dpi X 4096dpi or 2048dpi X 4096dpi, the detection accuracy will be greatly improved, and at the same time, the detection result will be more stable because of avoiding the ghost image at the edge.

Using testing equipment for quality inspection can provide real-time report and detailed and perfect analysis report of the whole inspection process. On-site operators can timely adjust the problems in their work according to the real-time analysis report with the help of the timely alarm of automatic detection equipment, which may reduce the rejection rate by more than one percentage point. Managers can track the production process according to the analysis report of the test results, which is more conducive to the management of production technology. Because customers require high-quality testing equipment, they should not only test the quality of printed matter, but also have the ability to analyze it afterwards. What some quality inspection equipment can do can not only improve the qualified rate of finished products, but also help manufacturers improve their technological processes, establish quality management systems and achieve long-term and stable quality standards.

Position Control and Product Inspection of Gravure Printing Machine

The video images of printed products are continuously shot by the camera set on the production line, and the camera speed is below 30 frames/second and adjustable. The image captured by the camera is first quantized, and the analog signal is converted into a digital signal, from which the key frames that effectively represent the lens content are extracted and displayed on the display. For a frame of image, we can use the analysis method of still image to deal with it. Through size measurement and multi-spectral analysis, we can identify the color scale on the video image, and obtain the color scale spacing, color parameters of the color scale and other correlations.

Due to various factors, there will be various noises, such as Gaussian noise, salt and pepper noise and random noise. Noise brings many difficulties to image processing, which directly affects image segmentation, feature extraction and image recognition, so it is necessary to filter the real-time collected images. Image filtering requires that the noise outside the image can be removed while maintaining the details of the image. When the noise is Gaussian noise, linear filter is often used, which is easy to analyze and realize. However, the linear filter has poor filtering effect on salt and pepper noise. The traditional median filtering can reduce the salt and pepper noise in the image, but the effect is not ideal, that is, the fully dispersed noise is removed and the noise close to each other is retained, so when the salt and pepper noise is serious, its filtering effect is obviously worse. The improved median filtering method of this system. This method firstly obtains the median after removing the maximum and minimum gray pixels in the noisy image window, then calculates the difference between the median and the gray value of the corresponding pixel, and then compares it with the threshold to determine whether to replace the gray value of the pixel with the obtained value.

At this stage of image segmentation, each color mark is detected and separated from the background. The edge of the object is reflected by the gray discontinuity. There are two kinds of L-shaped edges. One is the step edge, and the gray values of pixels on both sides of it are significantly different. The second is the roof-like edge, which is located at the turning point L where the gray value changes from increasing to decreasing. For step edge, its second directional derivative has zero crossing at the edge, so differential operator can be used as edge detection operator. The edge detection method of differential operator is similar to Qualcomm filtering in high-dimensional space domain, which can increase high-frequency components. This operator is quite sensitive to noise. For step edge, the commonly used operators are gradient operator, Sobel operator and Kirsh operator. Laplace transform and Kirsh operator can be used for roof-like edges. Because the color code is rectangular and the gray levels of adjacent edges are very different, edge detection is used to segment the image. In this paper, Sobert edge detector is used for edge detection. Using local difference operator to find the edge can better separate the color scale. In the actual detection process, the method of color image edge detection and appropriate color basis (such as intensity, chromaticity, saturation, etc.) are adopted. ) is selected for detection. According to the type characteristics of the printing press, that is, the color and layout characteristics of the printing press, multi-threshold processing is carried out to obtain the binary image of each color.

The segmented image is measured, and the object is identified by the measured value. Because color patches are rectangles with regular shapes, the following features can be extracted: (1) rectangular area is calculated by pixels, (2) squareness, (3) chroma (h) and saturation (s), and then the intervals between color patches are obtained according to the number of pixel points at each color patch interval, and compared with the set value, two are obtained. So as to eliminate or reduce printing dislocation. In feature extraction, multispectral image analysis of an image can quantitatively represent the color scale, such as the color of pixels in a color number image. Using HIS format to detect ink quality, two parameters of color information of each scale, chromaticity and saturation, are obtained. The binary image of each color is statistically calculated or matched with the standard image to measure the parameters such as ink chips in the printing process.

The printing machine is uncoiled by an uncoiler, printed and dried in various colors by each printing unit in turn, and wound by a winder. L For each color printing, a color mark for color registration is printed on the edge of the printed material. The level of the mark is 65,438+00 mm and the width is1mm.. When overprinting is accurate, the marking lines of each adjacent color should be parallel to each other, with a vertical (vertical) phase of 20 mm. The video image of printed matter is continuously shot by the camera set on the production line. Through size measurement and multi-spectral analysis, each color scale on the video image can be identified, and the color scale spacing and color parameter L of the color scale can be obtained. If the distance between two adjacent color codes is greater than or less than 20 mm, it means that there is a deviation in overprinter. The deviation signal is sent to the servo frequency conversion drive unit to drive the AC servo motor, so that the corresponding registration correction roller ML moves up and down, so as to prolong or shorten the stroke of the printed material from the previous unit plate roller to the unit plate roller for dynamic correction. In the automatic production of modern packaging industry, it involves all kinds of detection and measurement, such as printing quality detection of beverage bottle caps, bar code and character recognition on product packaging, etc. The same feature of this kind of application is continuous batch production, which requires very high appearance quality. Usually, this highly repetitive and intelligent work can only be completed by manual inspection. We often see hundreds of inspection workers performing this process behind the modern assembly lines of some factories. While adding huge labor costs and management costs to the factory, it is still impossible to guarantee the inspection qualification rate of 100% (that is, zero defects), and the competition among enterprises is no longer allowed. There is a defect of 1%. Sometimes, such as accurate and rapid measurement of micro-size, shape matching, color recognition and so on. It is impossible for human eyes to proceed continuously and stably, and other physical quantity sensors are also difficult to use. At this time, people began to consider the rapidity, reliability and repeatability of computers, thus introducing robot vision technology.

Generally speaking, firstly, the photographed object is converted into an image signal by a CCD camera and transmitted to a special image processing system. According to pixel distribution, brightness, color and other information, such as area, length, quantity, location, etc. Finally, output the results according to the preset tolerance and other conditions, such as size, angle, offset, quantity, qualified/unqualified, yes/no, etc. Machine vision is characterized by automation, objectivity, non-contact and high precision. Compared with general image processing systems, machine vision emphasizes accuracy and speed, as well as reliability in industrial field environment. Machine vision is very suitable for measurement, inspection and identification in mass production, such as identification of printed characters on IC surface, identification of production date on food packaging, and inspection of label placement position. In the machine vision system; The key technologies are light source lighting technology, optical lens, camera, image acquisition card, image processing card and fast and accurate actuator. In the machine vision application system; Good light source and lighting scheme are often the key to the success or failure of the whole system; Play a very important role; It doesn't simply illuminate objects. The matching of light source and lighting scheme should highlight the object characteristics as much as possible; The parts that need to be detected and the unimportant parts should be as obvious as possible; Increase contrast; At the same time, it is necessary to ensure sufficient overall brightness; The change of object position should not affect the imaging quality. Machine vision application systems generally use transmitted light and reflected light. In the case of reflected light, the relative position of light source and optical lens and the texture of the object surface should be fully considered. The geometry, background and other elements of an object. The choice of light source must conform to the required geometry, illumination brightness, uniformity, emission spectrum characteristics and so on. At the same time, the luminous efficiency and service life of the light source should be considered. Optical lens is equivalent to the lens of human eye; This is very important in machine vision system. The imaging quality of a lens is good or bad; That is, whether the aberration correction is good or not; Can be measured by aberration; Common aberrations are spherical aberration, coma aberration, astigmatism, field curvature, distortion and chromatic aberration.

Camera and image acquisition card * * * complete the collection and digitization of material images. High-quality image information is the original basis for correct judgment and decision-making of the system; This is another key to the success of the whole system. In the machine vision system; CCD camera is widely used because of its small size, reliable performance and high definition. According to the CCD devices used, CCD cameras can be divided into two categories: linear array and area array. Linear CCD camera can only get one line of image information at a time; The subject must move straight forward from the camera; So as to obtain a complete image; Therefore, it is very suitable for image detection of logistics moving at a certain speed; Area CCD camera can get the information of the whole image at one time. Image signal processing is the core of machine vision system; Equivalent to the human brain. How to process and calculate images; That is, the algorithms are all reflected here; It is the key and difficult point in the development of machine vision system. With the rapid development of computer technology, microelectronics technology and large-scale integrated circuit technology; So as to improve the real-time performance of the system; A lot of work of image processing can be done by hardware; Such as DSP, special image signal processing card, etc. The software mainly completes the part of the algorithm that is very complex and immature and needs to be explored and changed constantly.

From the product itself, machine vision will rely more and more on PC technology, and it will be more closely combined with other control and measurement such as data acquisition. And embedded products will gradually replace board products, which is a growing trend. The main reason is that with the rapid development of computer technology and microelectronics technology, the application field of embedded system is more and more extensive, especially its low power consumption technology has attracted people's attention. In addition, most embedded operating systems are based on C language, so it is a basic work to develop embedded systems using C high-level language. The advantage of using high-level language is that it can improve work efficiency and shorten the development cycle. More importantly, the developed products have high reliability, good maintainability and are convenient for continuous improvement and upgrading. Therefore, embedded products will replace board products.

Because machine vision is a part of automation, there is no machine vision without automation. Machine vision software and hardware products are gradually becoming the core system in different stages of collaborative manufacturing process. Both users and hardware suppliers regard machine vision products as tools for information collection on production lines, which requires a large number of standardization technologies for machine vision products. Intuitively speaking, it will gradually open up with the opening of automation, and can be re-developed according to the needs of users. Now automation enterprises are advocating the solution of software and hardware integration. In the next 5-6 years, machine vision manufacturers should not only be suppliers of products, but should gradually move towards system integrators of integrated solutions.

In the next few years, with the development of China's processing and manufacturing industry, the demand for machine vision will gradually increase; With the increase of machine vision products and the improvement of technology, the application of machine vision in China will shift from the initial low-end to the high-end. Due to the intervention of machine vision, automation will develop in a smarter and faster direction.