Overview of Machine Vision

	Machine vision	Artificial vision
Efficiency	High-efficiency	Low-efficiency
Speed	Fast	Slow
Stability	Stable inspection effect	Fatigable and emotional, hard to maintain the inspection effect
Working hours	Round-the-clock	Fatigable, limited working hour
Information integration	Achievable information integration	Hard to achieve information integration
Cost	One-time cost	High cost
Environment	Adaptable to inspection in dangerous environment	Unadaptable to inspection in dangerous environment

Comparison between machine vision and artificial vision      

In the past, artificial vision is the major method for the measurement, recognition and analysis in the process of the volume industrial production. Due to the low efficiency, poor stability and low accuracy of artificial vision processing, machine vision can greatly improve the processing efficiency and automation. Meanwhile, machine vision is also adopted to replace the artificial vision in hazardous working environments that are not suitable for manual operations or where the human eye cannot meet the requirements, including nuclear power plant monitoring and wafer defect detection. More importantly, machine vision can better facilitate information integration, which is one of the basic technologies for computer integration. Machine vision system can acquire massive information quickly and better facilitate the automation processing and information integration. Therefore, it is widely applied in the modern automatic production, including assembly positioning, product quality inspection, product recognition, dimension measurement, etc. The main difference between the machine and artificial vision is list in the table. 

History of machine vision development 

◆Development of machine vision in China and abroad   

In abroad, machine vision technology started earlier, the concept of machine vision appeared in the 1950s, and it really began to develop in the 1970s, and entered a period of rapid development in the late 1990s. There are two big leaps have been taken in the development of the machine vision. The first leap appeared in the 1970s at the innovation of CCD image sensor. The optics image can be transferred to the digital signal that processable for the computer, which marked the turning point in the development of machine vision. The second appeared in 1980s when CPU, DSP and other image processing technology achieved rapid development, which is also the advantageous condition for the development of the machine vision. Judging from the current layout of the global machine vision industry, industrial powers such as China, Germany, the United States, and Japan occupy most of the market for machine vision technology and applications. In abroad, machine vision is widely applied in semiconductor, electronic information, automotive, food, medical industries, etc. In the 21^st century, despite of the slowing down of the foreign machine vision market, it remains to play a bridge head role in the technology field.  In China, the machine vision started in the 1980s, kept on track from the late 20^th century to the early 21^st century, and on the rise after 2010. Along with improvement of industrial automation, requirements for the quality becomes higher, it becomes trendy that machine vision will take the place of artificial inspection. In addition, the industrial equipment in early China is featured with low automation. Therefore, the requirement for the upgrading strongly stimulates the massive market demand for machine vision. Along with the gradual development of the machine vision technology, many domestic technology companies are researching and promoting new products. Many R&D driven manufactures in China are emerging in the field of camera, shot, light source, image processing, etc. Affected by the combination of the demographic dividend of the manufacturing industry, the stimulation of smart manufacturing policies and the urgent need to improve factory automation, China has become one of the most promising and active regions in the development of machine vision around the world. 

◆Trend of machine vision 

With the complex and diverse application of machine vision, the combination with deep learning algorithms, 3D application technologies, interconnection standards and other technologies is becoming closer. Deep learning algorithm: this algorithm simulates the hierarchical structure of human brain, and realizes the hierarchical expression of data by constructing the low-level signal and high-level semantic. Deep learning algorithm is introduced to the machine vision image processing system for appearance inspection, making the recognition more intelligent and the information processing capability stronger. Application of 3D technology: with the development of the 3D application technology, more 3D reconstruction technology has been introduced to the machine vision, including structured light, DFF, TOF, binocular vision and photometric stereo, etc. 3D image processing and analysis algorithm has also been widely studied, which will become the mainstream for the development of machine vision. Standard for interconnection: it is necessary to realize the interconnection among the machine vision system, intelligent manufacturing equipment and the management system of enterprises so that equipment and manufacturing management will become more intelligent. Currently, EMVA, AIA, CMVA and JIIA jointly form the GenICam for the machine vision industry. AIA formed GigE Vision, USB3 Vision and other camera communication protocols. Machine vision industry has also expanded its boundary through the cooperation with other industries, aiming to realize the combination of the vision system with other industries.