"AI+Machine Vision", a new horizon in the field of intelligent machine vision

Jiwei Network reported that from manufacturing to "intelligent manufacturing", digital transformation is deepening into major industrial fields. From machine interconnection to human-machine collaboration to unmanned factories, under the global tide of industrial transformation and manufacturing upgrading, every factory and every enterprise is promoted to "embrace" digitalization.

So what exactly does the real digital transformation upgrade have to do? Replacing the "human eye" with machine vision is an important step towards industrial intelligence. Relevant statistics show that more than 75% of the information obtained by people comes from vision, machine vision is the re-improvement of human eye vision, and visual perception is also an important data entry for industrial manufacturing to achieve deep digital transformation.

The domestic manufacturer Xinge Intelligence is committed to improving and exploring the core software and hardware capabilities of machine vision and industrial application scenarios. Deep ploughing of the machine vision track for 5 years, Xinge Intelligent has grown rapidly, and is currently the only product company in China that has the core technical barriers of the whole industry chain in the field of machine vision and optical measurement, and its technical capabilities cover chip level, product level and program level. Its products have been widely used in various industries in the industrial field, including semiconductors, 3C, automotive, lithium batteries, welding and so on.

Recently, Xinge Intelligence has launched a series of updates on related capabilities around machine vision, of which the combination of artificial intelligence (AI) and machine vision algorithms is a key highlight. Xinge Intelligent pointed out to Jiwei Network that continuously improving the ability and accuracy of machine vision through machine learning and artificial intelligence optimization is the focus of the company's next efforts. The more stable application of artificial intelligence in the industrial environment is also of great significance to improve the level of intelligence of overall industrial manufacturing.

AI+ machine vision to create industrial intelligence

Detection, judgment, identification and control in industrial manufacturing are collectively referred to as "machine vision". In recent years, under the trend of transformation of manufacturing to automation and intelligent upgrading, machine vision has become one of the most important and widely used fields, with a market prospect of 100 billion yuan. One of the cores of smart manufacturing is the machine vision system. Compared with the human eye, machine vision has significant advantages such as high precision, fast speed and strong adaptability. Machine vision instead of the human eye can achieve a variety of functions in a variety of scenarios, and is widely used in intelligent manufacturing, high-end equipment, automation equipment and other industries.

At present, many factory equipment has achieved basic automation, which is manifested as a mechanical repetitive action that can be manually programmed. However, in the field of industrial testing, the enterprise has completed the project engineering problem modeling and intelligent system development, as well as the key tool parameters offline measurement and online soft measurement, although the data provided basically meets the digital manufacturing represented by the data kanban, but it is not easy to provide high-quality data on the spot, let alone the intelligence with the ability to actively perceive and analyze and think decision-making, so the machine is only the automation level. After the introduction of AI vision, the engineering problems and engineering parameters can first be modeled, and then the machine learning of the model can be deeply bound to the mechanical equipment and production status by using the collected high-quality data, and then the intelligent system can be developed on the basis of which the intelligent system is developed, and then the production parameters that are instantly variable and can be maintained and optimized are finally handed over to the basic automation to achieve a comprehensive upgrade of mechanization- automation- digitization- and intelligence.

However, at present, in the industrial field, the attempt to integrate AI into machine vision algorithms has just started in the industrial field, and the world is still in the initial stage of exploration, and a large number of mature applications have not really appeared. However, many domestic and foreign companies have begun to delve into industrial application scenarios and actively develop the best application solutions for industrial AI.

The path of Xinge's intelligent exploration is to pick out the part of the AI algorithm that is most suitable for integration with machine vision, and combine the two into one to continuously optimize the application of machine vision intelligence. According to reports, the industrial AI algorithm is mainly used to include: machine vision product detection, such as appearance, quality, size, color, grading and other detection, mainly for pharmaceutical manufacturing, 3C products, precision devices, flooring, cloth, etc.; personnel behavior detection, for such as abnormal behavior, operation specifications, safety, artificial efficacy monitoring, etc., mainly for intelligent production lines, factory behavior identification, etc., such as engineering construction, factory production, ship shipping, farm food and other industries; equipment failure model detection, It is used for such as equipment, assembly line failure model prediction, etc., mainly for large machine tools, assembly lines, construction machinery and other fields; OCR recognition, such as character and text recognition, mainly for assembly line work, material storage center, product information collection center, etc.

Figure: CoreGe Intelligent laser 3D contour camera products

Figure: CoreGe Intelligent Laser Displacement Sensor Products

According to Xinge Intelligence, in the algorithm design, the factors of machine vision and machine learning are considered at the same time, and redundancy and reuse are avoided, which is conducive to ensuring the efficiency of algorithm operation. It is worth mentioning that in order to overcome the problem of algorithm training caused by the lack of defect samples in the field of industrial quality inspection, Xinge Intelligent adopts a leading small sample size learning method to develop the algorithm, which can eliminate the need for complex sampling preparation in practical applications. It is reported that the company's current visual algorithm system includes mainstream algorithms such as deep learning, traditional comparison algorithms, 3D point clouds, etc., which can cover more than 90% of conventional applications. Compared with the traditional machine vision algorithm, the machine vision algorithm with artificial intelligence can improve the overall detection speed and accuracy through model, algorithm and instruction optimization, and perfectly adapt to changes in rotation, indentation, translation, chromatic aberration, light intensity, etc., to achieve accurate detection of flexible objects and partially occluded objects.

Illustration: Comparison of traditional machine vision and deep learning applications

In the solution of integrating AI into machine vision, thanks to the core technical capabilities of chip-level, product-level and program-level, Xinge Intelligent can achieve the best platform for algorithm implementation through the selection and construction of chips and solutions when creating artificial intelligence algorithms. The overall solution can include algorithms, equipment, programs, and even the entire production system (such as loading system, optical imaging system, unloading system, AI discrimination system, man-machine interface), and can flexibly adjust the program delivery mode, which can be delivered as a whole or spun off, or can be fully customized according to customer needs.

Intelligent machine vision is the golden track The local market space is huge

As the world's factory, China has a large number of application scenarios. The current industrial automation market has bred a number of very powerful domestic vision solution companies, which have replaced and upgraded a considerable number of import solutions, but at present, the penetration rate of machine vision core software and hardware is still low, and there is still a lot of room for improvement.

In the view of Liu Jian, founder of Xinge Intelligence, the current machine vision industry is still in a state of "race". Compared with the global machine vision industry, the domestic machine vision industry started late. Roughly reviewing the development process of the industry, 1999-2003 is the enlightenment stage of domestic machine vision development, and cross-professional machine vision talents began to appear; after 2004, into the early stage of development, machine vision companies began to explore and develop independent technologies and products, and made certain breakthroughs. Since then, after more than ten years, China's machine vision industry has gradually moved towards a stage of rapid development.

For the main bottleneck of the current overall machine vision application penetration rate is not high, The core song intelligent analysis believes that, first of all, from the perspective of application reality, the corresponding cost of products and solutions is still high; secondly, the implementation of the program to the landing has a certain degree of difficulty, because the industrial application scenario is very fragmented, and involves a large number of industries Know-how, and the requirements for practitioners are also very high. In addition, the overall machine vision industry is still the leading level of foreign enterprises. According to the statistics of China Machine Vision Industry Alliance, the current mainland machine vision enterprises are generally small and medium-sized enterprises, with sales of more than 100 million yuan only 16.5%, nearly 20% of enterprise sales less than 10 million yuan.

However, Xinge Intelligence pointed out that intelligence is a global topic, and artificial intelligence, machine vision and other fields are areas of rapid development in the world, which is a global track. In terms of artificial intelligence or deep learning machine vision, China is not in a backward state, and even the impact of the epidemic on overseas industrial fields is currently in a relatively leading position in China. At present, in the exploration and application of industrial AI, there is no particularly mature and large-scale solution system in the world.

Xinge Intelligence foresees that industrial AI+ machine vision will be the golden track for sustainable development in the future, and the next level of competition, continuous technological improvement is on the one hand, and the ability to apply landing is more critical. The fragmentation of application scenarios is the main pain point that plagues the large-scale landing of intelligent manufacturing-related applications. In this regard, Xinge Intelligent pointed out that in the long run, it is necessary to complete successful cases of industry leaders and form products and programs accordingly. More importantly, it is necessary to build a more comprehensive, effective and active industrial "standard" system for multiple industries to accelerate the landing and large-scale promotion of industrial AI. In recent years, Xinge Intelligence is also actively trying to modularize and standardize the path to promote the landing of intelligent machine vision programs, through the continuous creation of successful cases in different industries, The Xinge intelligent team in this process is also constantly accumulating the modularization and standardization of algorithms and corresponding controllers, while improving the technical and program capabilities, but also accumulating industry know-how, is conducive to the application and deployment of intelligent machine vision solutions in various vertical fields.

As the world's largest manufacturing base, China is also becoming an important application market in the global vision field, and multiple application areas promote the machine vision industry to maintain high prosperity and high growth. At present, consumer electronics and automobiles are the largest application areas of machine vision in China, and pharmaceutical and food packaging are also potentially huge application areas of machine vision. Especially in recent years, the electrification and intelligence of automobiles have become another important driving force for machine vision, and the rapid development of new energy vehicles in recent years has led to the rapid development of the entire automotive electronics industry chain. Due to the significant increase in precision parts of intelligent electric vehicles, higher requirements are put forward for the efficiency and accuracy of detection; under the development trend of automobile intelligence and lightweight, the demand for machine vision in the automotive industry is expected to continue to increase.

In fact, in recent years, local manufacturers have begun a comprehensive exploration based on the actual needs of the domestic market, starting from various subdivisions, and a large number of fast-growing local machine vision industry chain enterprises have emerged, and have begun to form autonomous breakthroughs. In the long run, the aging of the population and the improvement of labor costs will endogenously drive the continuous growth of the machine vision industry, and the industrial policies of various intelligent manufacturing branches will continue to promote the localization of machine vision. This is a rare development opportunity for local manufacturers.