Maschinelles Sehen wird auf die Halbleitererkennung angewendet, und die Jushi-Technologie fördert die Anwendung der KI-Erkennung in industriellen Umgebungen

Es versteht sich, dass das Xilang Institute und einige seiner Inkubatoren vom 9. bis 11. September 2020 eine große Ausstellung am Stand 1B41 in Halle 1 der 22. China International Optoelectronic Expo veranstalten werden. Wir laden Sie herzlich ein, uns zu besuchen, zu kommunizieren und Geschäfte zu verhandeln. Den QR-Code über dem Ausweis lange drücken, sofort anmelden und die Besuchsbescheinigung kostenlos erhalten.

In den letzten Jahren standen chinesische Halbleiterunternehmen im Fokus der Vereinigten Staaten. Im Jahr 2019 hat das Halbleiter-Umsatzvolumen auf dem chinesischen Festland 40 Milliarden US-Dollar überschritten. Die globale Halbleiterindustrie verlagert sich nach und nach auf das chinesische Festland, und das Festland entwickelt sich zunehmend zu einer neuen Halbleiterbasis. Andererseits sind unsere Halbleiterprodukte stark von Importen abhängig. Nach Angaben der General Administration of Customs erreichte Chinas Importumfang von computerintegrierten Technologieprodukten im Jahr 2017 260,1 Milliarden US-Dollar (etwa 1.756,1 Milliarden Yuan), was 14,1 % des gesamten Importvolumens Chinas in diesem Jahr ausmachte, und das Importvolumen überstieg 200 Milliarden US-Dollar Dollar für vier aufeinanderfolgende Jahre.

Aus der Perspektive des externen Umfelds, derzeit ausgesetzt Intel, Qualcomm, Celinx, Broadcom und andere internationale Chiphersteller den Handel, Google Handy-Betriebssystem Android, Microsoft Windows ausgesetzt, um neue Aufträge zu erhalten, wurde das Risiko eines Technologie-Embargos vollständig ausgelegt Vor Menschen. Um die große Wiederbelebung der chinesischen Nation zu verwirklichen, muss die Kerntechnologie eigenständig sein, wobei die mit Halbleiterchips verbundenen Bereiche der Technologieautonomie und -kontrolle zum Schlüssel dafür werden, ob das Ziel erreicht werden kann.

From the perspective of internal industry development, the current semiconductor industry is divided into the front (wafer production), the middle (wafer manufacturing) and the back (wafer packaging and testing), among which semiconductor detection equipment is the key to chip yield control, throughout the entire life cycle of high-end semiconductor manufacturing, to some extent can be said to be the "rigid need", is the core of the core.

It is no exaggeration to say that the entire semiconductor manufacturing industry itself is a quality control process.

There are tens of millions of circuits on the chip of square inch size, so the semiconductor manufacturing has a very high requirement for detection accuracy. In this context, the force that artificial eyes can play is extremely limited, and the whole industry chain is almost completely dependent on machine vision. According to gartner's data, in 2019, the global wafer processing market scale is about $62.7 billion, accounting for nearly 20% of the global semiconductor market, and the compound growth rate of wafer processing market is expected to reach 4.9% from 2018 to 2023. With the shift of the global semiconductor manufacturing focus to the domestic market, the global market share of China's wafer processing, sealing and testing market has been increasing year by year. By October 2018, the total equipment investment of 17 fabs built in China has reached 500 billion yuan, and the market space of forward volume testing equipment will reach 46 billion yuan. It is also expected to drive the demand of 45 billion yuan of back-end detection equipment.

Such a huge market, but to face the lack of independent technology.

For a long time, the semiconductor testing field in China has been almost monopolized by foreign equipment manufacturers such as Teradyne, Xcerra and Advantest, but this situation is being broken by a startup company from China.

Jushi Technology (www.matrixtime.com) is a local Chinese machine vision manufacturer, aiming to provide industrial inspection solutions for high-end manufacturing scenarios such as semiconductors with artificial intelligence technology.

As a hard technology AI company deeply engaged in deep learning and computer vision, Jushi Technology is committed to solving difficult scene problems in manufacturing industry with AI technology. The full deep learning-driven machine vision inspection system developed by the company has been successfully applied to defect detection and quality control in the field of semiconductor back-channel, and has been effectively verified in customer production. Currently, Jushi's technology can examine the internal structure of a 14-nanometer chip, the equivalent of looking at a nail on Earth from space.

In terms of core technology, Jushi applies AI technology and algorithm to industrial intelligent manufacturing by deep learning, computer vision, motion planning and other core technology research and development, and develops scenario-based industrial AI products. The solution of Jushi Technology is highly productized and agile project development delivery mechanism. At the present stage, Jishi Technology has two major products: MatrixSemi® for AI visual inspection and MatrixRobot® for industrial robot "eyes". Starting from customization business, Jishi Technology goes deep into vertical industries and gradually forms relatively standardized products. Based on this, Jishi Technology customises end-to-end technical solutions for various industrial scenarios.

Taking the current application in the field of semiconductor back-channel detection as an example, currently such as insufficient corrosion, over corrosion, sand holes, lack of silver, plating offset and other frequently occurring dozens of product defects in the semiconductor manufacturing industry, JUS Technology has launched MatrixSemi equipment-level solutions. This solution is a "full deep learning driven" semiconductor defect AI detection platform specially launched by JUS Technology for semiconductors, including special lighting imaging, AI detection algorithm, underlying acceleration, ADC system and other functions. Special lighting imaging can provide micron, submicron level dedicated lighting imaging; Semiconductor dedicated deep learning model, including dozens of deep learning models, 3D/2D visual models and algorithms, specific visual algorithms and quality control rule parsing engine. At the same time, the solution incorporates the latest full-stack AI acceleration technology to support semiconductor high-resolution visual real-time edge computing, reducing CT time. Automatic machine learning defect analysis technology ADC is provided to achieve closed-loop semiconductor quality management. Based on this core technology, the semiconductor AOI (automatic optical detection) system independently developed by JUS Technology -- JUS 2000 is the first semiconductor AI visual detection system equipment at home and abroad. It covers more than 30 deep learning model algorithms, and the detection accuracy has been subsubversive improved, which is 10 times higher than the current foreign system. Under the condition that the missed detection rate is 0, The actual false detection rate of the detection system is maintained between 1%-5%, and it can also achieve specific classification, visual positioning and quantitative detection of defect links, and set different detection standards for different customers according to different defects. In addition, for different product testing, core 2000 can also automatically adapt and test standard changes. Currently, it is mainly used in semiconductor advanced manufacturing and complex 3C precision manufacturing defect detection, including front/back channel, wafer related process, lead frame stamping and etching, LED chip detection, SiP packaging detection, complex 3C manufacturing and LCD panel detection and other scenarios.

"Using deep learning and complex machine vision to improve quality detection and control is of great significance for high-end semiconductor manufacturing," said Zheng Jun, founder and CEO of Jushi Technology. "The more complex the scene is, the more obvious our AI detection advantage is. In the field of industrial machine vision, We have surpassed traditional monopolies such as Cognex of the United States, Keenes of Japan and Germany in complex scenarios and won a large number of orders from well-known customers, which proves that Core 2000 is globally competitive."

In addition, compared with the leading technology level, the advantage of Jushi technology is that its products have extremely high adaptability and convenience.

As semiconductor design and manufacturing become more specialized, nearly every wafer factory requires dozens of different large equipment to complete hundreds of steps in the wafer production process, which makes it difficult to add any one process flow, especially for universal equipment in the inspection process, often requires complex modifications to the production line. And long hours of training for employees.

The Jushi 2000, on the other hand, can be used in a one-step "foolproof" way using an AI system -- workers or engineers can quickly pick up and use it with little training. On the other hand, the unique embedded design of Juji Technology enables the finished products and defective products to be clearly separated at the discharge end once the core 2000 is connected to the normal production line, without the need to send the products to the testing room and laboratory separately for testing. This alone can reduce the testing cost for enterprises and integrators by more than 10 times.

This ability to commercialize advanced technologies through engineering capabilities and product thinking is the core competitive barrier of Jushi technology. At present, Core 2000 has launched the production line of Ningbo Kangqiang Electronics, the largest semiconductor lead frame manufacturer in China, in September 2019. The operation effect is good, and has been widely praised by the enterprise.

From the perspective of the team, Jushi Technology can also be called "core" member configuration. The company does not even have direct sales staff. More than 90% of the team are product R&D and technical staff, including 22 PHDS, and there are many technical experts from Siemens, Foxconn and other specific production lines. Zheng Jun, the founder and CEO, used to work at Bell LABS and had experience in Foxconn, SAIC and other B-terminal production lines.

At present, many start-up companies have applied machine vision to industrial inspection, but most of them will choose to cut into machine vision inspection from the relatively easy 3C and auto parts industries. Only Jushi Technology chose the most difficult AI application, semiconductor detection known for high precision as a breakthrough.

Everything seems to be just as the founder Zheng Jun said, "Use the most interesting technology to do the most meaningful things, often do the most difficult things is the most valuable."

Currently, JUS has more than 40 patents in depth fields such as deep learning and machine learning, computer 2D/3D vision, computer graphics, lighting imaging and intelligent machine control.

Durch fortschrittliche Technologie und unabhängige Forschung und Entwicklung hat Jushi Technology erfolgreich eine Innovationslücke im Bereich der Chiperkennung geöffnet, die von ausländischen Giganten umgeben ist, und eine vorteilhafte Erkundung für die inländische High-End-Halbleiterherstellung durchgeführt. Zheng Jun sagte, dass das Unternehmen in diesem Jahr die Produktforschung und -entwicklung des Algorithmus sowie die Marktreplikation und Geschäftsexpansion abschließen werde. Im Jahr 2020 wird sich Jushi Technology auf drei Schaltkreise konzentrieren: Halbleiter-, Photovoltaik- und Automobil-Präzisionsfertigung.