With the continuous adjustment of the global manufacturing pattern, the traditional manufacturing industry is deeply integrated with the new generation of technology. Machine vision technology helps the continuous upgrading and transformation of the manufacturing industry, and to a certain extent promotes the development of intelligent manufacturing.

First, the basic functions of machine vision and the main application industries

First of all, industrial machine vision can realize the following four basic functions, namely, identification, measurement, positioning and detection:

A. identification. It mainly identifies the physical features of the object, including surface features such as shape, color, character and bar code. The main indexes to measure the recognition accuracy are accuracy and speed recognition. It is mainly used in the traceability of materials, processes and stations. The method is to read letters, numbers and characters (such as bar codes, two-dimensional codes, etc.) on parts.

B. measure. By calibrating the obtained image pixel information as the measurement unit, the geometric size of the target object is accurately calculated, which is mainly used in high-precision and complex morphometry.

C. positioning. It refers to the acquisition of the spatial position information of the target object, which can be divided into two-dimensional or three-dimensional information. It is mainly used to complete the auxiliary operation, which is often used to assist the assembly and picking of the robot.

D. detection. Is to detect the surface state of the target object, so as to determine whether the product has quality defects, such as parts appearance defects, contaminants attachment, functional defects, etc.

Secondly, industrial machine vision is mainly applied in the following industries:

A. Consumer electronics industry. The application of this industry is mainly reflected in the main board, parts assembly, machine assembly of these three major production links. As consumer electronics become more sophisticated, quality standards are rising as components get smaller and smaller. Therefore, the demand for industrial machine vision is constantly enlarged. Taking 5G smart phones as an example, product upgrading and technology upgrading correspondingly require machine vision tools to be upgraded. In the motherboard and parts assembly, still to 2D vision, 3D vision as a supplement. In the assembly of the whole machine, the main is still manpower. Machine vision mainly do appearance detection, the most is to do glass detection. In defect detection, machine vision is the most widely used place. Its high precision, high speed detection ability, can be very good to complete the scratch, damage, spot, color difference detection.

B. Semiconductor industry. The semiconductor industry is the earliest and more mature field of industrial machine vision application, which is also related to the rapid iteration and upgrading of the semiconductor industry. Its high-end market is basically occupied by overseas manufacturers. On the other hand, it also has something to do with the fact that the precision of semiconductors is so high that manual testing cannot play its proper role. For example, the appearance defects, size, quantity, flatness, distance, positioning, calibration, solder joint quality, bending and other tests of semiconductors, especially the detection, positioning, cutting and packaging of chips need to be dominated by industrial machine vision. Taking cutting as an example, rapid and accurate positioning is required. If the location goes wrong, the entire chip will die. The whole cutting process also needs the machine vision system for the whole positioning guidance. After cutting is completed, non-defective products are identified by machine vision and entered into the patch process.

[C]. The automobile industry. Today's automobile industry has achieved a high degree of automation, industrial machine vision plays a huge role in production efficiency, quality assurance, safety and reliability. Machine vision has been throughout the entire automobile manufacturing process, including the initial raw material quality inspection to 100% online measurement of auto parts, and then the welding, gluing, punching and other technological processes to control, and finally the body assembly, vehicle quality control. In addition, the vision guidance technology guides the robot for the best matching installation, accurate hole making, weld guidance and tracking, spraying guidance, windshield loading guidance, etc. This is the main application area of the automotive industry and the main area of innovation for domestic companies at present. With the increasing proportion of electronic parts of new energy and intelligent vehicles, the role of industrial machine vision is becoming more and more important.

Two, the development trend of machine vision

Data show that the industrial machine vision technology market size has reached $4.44 billion in 2018, and is expected to reach $12.29 billion in 2023, at a CAGR of up to 21%. At present, machine vision is changing from traditional industrial vision to deep learning industrial vision, and the application field of industrial machine vision is becoming wider and wider in the future. Based on this, industrial machine vision can be regarded as an important branch of AI. In the future, industrial machine vision and AI will be combined to solve the problems of diverse image and video scenes, various objects, and uncontrolled conditions, where the object is affected by multiple variables such as illumination, pose, occlusion, etc., and to independently face depth scenes such as huge data volume, complex features, and real-time autonomous processing for some applications.

From the perspective of the current situation, there are still the following constraints: first, the cost of end-to-side computing power is getting higher and higher; Second, the maintenance cost of single point system is too high; Third, data islands; Fourth, versatility, intelligence and so on.

Three, the problems of machine vision talent training

Through investigation and analysis, the current machine vision talent training mainly has the following problems:

A. Machine vision is an interdisciplinary subject, students only master the relevant knowledge of this course, but the overall understanding of machine vision system is not enough;

B. The course teaching is separated from the practical application, and the knowledge learned in class only stays at the theoretical level, which on the one hand will lead to the incomprehension and indigestion of theoretical knowledge, and on the other hand can not effectively solve practical problems;

C. relatively old-fashioned classroom teaching, classroom teaching by correlation algorithm is the most basic, but with the rapid development of the neural network, and deep learning, machine vision should extend to a more intelligent, multidimensional, flexibility in the direction of development, therefore also calls for training students have certain thinking of artificial intelligence, and combined with machine vision technology.

Based on the above three problems, the school continues to encourage students to participate in skills practice and work positions, which can help students deepen their understanding of classroom knowledge and master professional skills.