Machine vision uses computers to simulate human visual functions, enabling machines to acquire and understand relevant visual information. It can be divided into two principles: "vision" and "perception."
"Vision" involves displaying external information as digital signals through imaging and feeding them back to the computer. This requires a complete hardware solution, including light sources, cameras, image acquisition cards, and vision sensors. "Perception," on the other hand, involves the computer processing and analyzing these digital signals, primarily through software algorithms.
Machine vision has a wide range of applications in industry, and its core functions include measurement, detection, identification, and positioning.
The industry chain can be divided into the upstream component market, the midstream system integration/complete equipment market, and the downstream application market.
The upstream of machine vision includes hardware and software providers such as light sources, lenses, industrial cameras, image acquisition cards, and image processing software. The midstream includes integrators and complete equipment providers. The downstream applications are extensive, with major downstream markets including electronics manufacturing, automotive, printing and packaging, tobacco, agriculture, pharmaceuticals, textiles, and transportation.
The global machine vision market is mainly distributed in North America, Europe, Japan, and China. According to statistics, the global market size for machine vision systems and components was US$3.67 billion in 2014, US$4.2 billion in 2015, and US$6.2 billion in 2016, with a compound annual growth rate of approximately 12% from 2002 to 2016. Machine vision system integration, based on North American market data, is estimated to be about six times the size of the vision systems and components market.
China's machine vision industry began with the introduction of technology in the 1980s. With the introduction of complete production lines in semiconductor factories in 1998, machine vision systems were also brought in. Before 2006, domestic machine vision products were mainly concentrated in foreign-funded manufacturing enterprises, which were all relatively small in scale. Starting in 2006, the customer base for industrial machine vision applications began to expand to the inspection fields of printing, food, etc. The market began to grow rapidly in 2011. With the increase in labor costs and the demand for upgrading in the manufacturing industry, coupled with the rapid development of computer vision technology, more and more machine vision solutions have penetrated into various fields. By 2016, the market size of my country's machine vision industry had reached nearly 7 billion yuan.
In machine vision, defect detection is one of the most widely used functions, primarily detecting various information on the surface of products. In modern industrial automated production, each process in continuous mass production has a certain defect rate. Although these rates may seem small individually, they multiply to become a bottleneck for companies to improve yield. Furthermore, removing defective products after the entire process is completed is much more costly (for example, if there is a positioning deviation in the solder paste printing process, and this problem is only discovered during online testing after chip mounting, the cost of rework will be more than 100 times the original cost). Therefore, timely detection and defect removal are crucial for quality and cost control, and are also an important cornerstone for the further upgrading of the manufacturing industry.
1. In the inspection industry, machine vision has significant advantages over human vision.
1) High precision: Human vision has 64 gray levels and is weak in distinguishing tiny targets; machine vision can significantly improve the gray level and can observe targets at the micrometer level.
2) High speed: Humans cannot see fast-moving targets clearly, while machine shutter speeds can reach the microsecond level;
3) High stability: Machine vision solves a very serious problem for humans: instability. Manual visual inspection is a very tedious and arduous job, and no matter what reward or punishment system you design, a relatively high rate of missed detections will occur. However, machine vision inspection equipment does not have fatigue issues or emotional fluctuations. As long as you write something in the algorithm, it will execute it diligently every time. This greatly improves the controllability of results in quality control.
4) Information integration and retention: The amount of information obtained by machine vision is comprehensive and traceable, and relevant information can be easily integrated and retained.
2. Machine vision technology has developed rapidly in recent years.
1) Image acquisition technology is developing rapidly.
With increasingly mature firmware for CCD and CMOS sensors, the size of image-sensitive devices continues to shrink, the number of pixels and data rates are constantly increasing, and the speed of improvement in resolution and frame rate can be described as rapid. Product series are also becoming more and more abundant, with continuous optimization of parameters such as gain, shutter speed and signal-to-noise ratio. By using core test indicators (MTF, distortion, signal-to-noise ratio, light source brightness, uniformity, color temperature, and comprehensive evaluation of system imaging capabilities, etc.) to comprehensively select light sources, lenses and cameras, many previously difficult imaging problems have been continuously overcome.
2) Image processing and pattern recognition are developing rapidly.
In image processing, with the extraction of high-precision edge information from images, many low-contrast defects that were originally difficult to detect directly due to being mixed in with background noise have begun to be distinguished.
In pattern recognition, it can be viewed as a labeling process, classifying the patterns to be recognized into their respective categories based on a certain amount of measurement or observation. Decision theory and structural methods are frequently used in image recognition. Decision theory is based on decision functions, which are used to classify and recognize pattern vectors, relying on time-dependent descriptions (such as statistical textures). Structural methods decompose objects into patterns or pattern primitives, with different object structures having different primitive strings (or strings). By using given pattern primitives to find the encoding boundaries of unknown objects, strings are obtained, and their class is determined based on these strings. In feature generation, many new algorithms are constantly emerging, including wavelet, wavelet packet, and fractal-based features, as well as unique binary component analysis; further developments include support vector machines, deformable template matching, and the design of linear and nonlinear classifiers.
3) Breakthroughs brought about by deep learning
Traditional machine learning relies primarily on human analysis and logic for feature extraction. Deep learning, however, simulates brain function through multilayer perceptrons, constructing deep neural networks (such as convolutional neural networks) to learn simple features, build complex features, learn mappings, and output results. During training, all layers are continuously optimized. Specific applications include automatic ROI segmentation; punctuation localization (allowing flexible detection of unknown defects through simulated vision); re-detecting indescribable or quantifiable defects such as orange peel flaws from heavily noisy images; and distinguishing between real and fake defects in glass cover detection. As more and more deep learning-based machine vision software enters the market (including Vidi from Switzerland, SUALAB from South Korea, and the Hong Kong Applied Science and Technology Research Institute), the empowering effect of deep learning on machine vision will become increasingly evident.
4) The development of 3D vision
3D vision is still in its early stages. Many applications are using 3D surface reconstruction, including navigation, industrial inspection, reverse engineering, surveying, object recognition, measurement and classification, etc. However, accuracy issues limit the application of 3D vision in many scenarios. Currently, the most widely used application in engineering is the volume measurement of standard parts in logistics, and it is believed that this area has huge potential in the future.
3. To completely replace manual visual inspection, machine vision still faces many challenges that need to be overcome.
1) Light source and imaging: High-quality imaging is the first step in machine vision. Since the reflection and refraction of different materials and surfaces can affect the extraction of features of the object being measured, the light source and imaging can be said to be the first major hurdle that machine vision inspection needs to overcome. For example, in the current detection of scratches on glass and reflective surfaces, the problem often lies in the integrated imaging of different defects.
2) Feature extraction in low-contrast images with heavy noise: In noisy environments, it is often difficult to distinguish between real and fake defects, which is why there is always a certain false detection rate in many scenarios. However, thanks to the rapid development of imaging and edge feature extraction, breakthroughs have been made in this area.
3) Identification of unexpected defects: In applications, specific defect patterns are often given, and machine vision is used to identify whether they have occurred. However, it is common to encounter situations where many obvious defects are missed because they have never occurred before or the patterns of occurrence are too diverse. If it were a human, even if they were not instructed to detect the defect in the operating procedure, they would notice it and have a higher probability of catching it. The "intelligence" of machine vision in this respect is currently difficult to achieve.
4. Machine Vision Industry Chain Status
1) Upstream component market
This primarily includes providers of light sources, lenses, industrial cameras, image acquisition cards, and image processing software. In recent years, smart cameras, industrial cameras, light sources, and circuit boards have all maintained a growth rate of no less than 20%. According to a survey by the China Machine Vision Industry Alliance (CMVU), there are now nearly 200 international machine vision brands in China (such as core component manufacturers like Cognex, Dalsar, and Baumer, and companies like Keyence, Omron, Panasonic, Banner, and NI that are involved in both core machine vision components and system integration). There are also more than 100 domestic machine vision brands (such as Hikvision, Huarui, Mengtuo Optoelectronics, Shenzhou Vision, Shenzhen Canrui, Shanghai Fangcheng, and Shanghai Bocheng Electric), and more than 300 distributors of various machine vision products (such as Shenzhen Hongfu Vision, Microvision New Era, Sanbao Xingye, Lingyun Optoelectronics, and Sunshine Vision). Many domestic machine vision component markets started as distributors of foreign brands. Many companies have established strong, often exclusive, partnerships with their international counterparts, creating barriers to entry for potential entrants. Consequently, distributors of high-quality products enjoy strong market competitiveness and profitability. Meanwhile, domestically produced industrial vision core components, represented by Hikvision and Huarui, are rapidly emerging.
2) Midstream system integration and complete equipment market
There are over 100 domestic midstream system integrators and complete equipment manufacturers that can provide comprehensive machine vision solutions to automation companies across various industries. These include companies like Lingyun Optoelectronics, Microvision New Era, Jiaheng, ADLINK, Sunshine Vision, Dingxin, and Daheng Imaging. However, due to the significant gap between domestic and international products, and because many midstream system integrators and complete equipment manufacturers started as traders of core components, they still tend to favor foreign brands when selecting vision products. To promote their hardware and software products, domestic brands often need to develop their own solution integration capabilities to better compete in the market.
3) Downstream application market
Downstream in the machine vision industry primarily consists of companies providing non-standard automation solutions to end users. This sector is highly industry-specific, and its core competitiveness lies in its comprehensive understanding of the industry and production processes, as well as the integration of various technologies. Because automation upgrades in various industries are cyclical and heavily influenced by the overall industry upgrade speed, shipment volume, and profit margins, the primary driver of machine vision application adoption in the past two years has been the electronics manufacturing industry, followed by the automotive and pharmaceutical sectors.
i. Semiconductor and Electronics Manufacturing Industry: Looking at the application distribution of machine vision in domestic industries, 46% is concentrated in the electronics and semiconductor manufacturing sector, including wafer processing and dicing, PCB inspection (film, inner/outer layer boards, final product appearance inspection, etc.), SMT assembly inspection, AOI defect detection throughout the LCD process, surface defect detection of various 3C components, and appearance inspection of 3C products.
ii. Automotive: Body assembly inspection, measurement of part geometry and errors, inspection of surface and internal defects, gap inspection, etc.
iii. Printing and Packaging Inspection: Tobacco shell printing, food packaging and printing, pharmaceutical aluminum-plastic composite packaging and printing, etc.
iv. Agriculture: Grading, inspection, and classification of agricultural products
v. Textiles: Detection of defects such as foreign fibers, cloud weave, warp defects, and weft defects; identification of fabric surface fuzz; yarn structure analysis, etc.
5. Future Development Trends of Machine Vision Systems
1) Embedded solutions are developing rapidly, smart cameras have outstanding performance and cost advantages, and embedded PCs will become increasingly powerful.
2) Modular, general-purpose software platforms and artificial intelligence software platforms will reduce the technical requirements for developers and shorten the development cycle.
3) 3D vision will be applied to more scenarios.
Disclaimer: This article is a reprint. If it involves copyright issues, please contact us promptly for deletion (QQ: 2737591964). We apologize for any inconvenience.
