——Interview with Li Hongkun, co-founder and CTO of Beijing Migration Technology Co., Ltd
Based on years of deep technical cultivation in the direction of 3D camera hardware, algorithms and software, Migration Technology has created a stable, easy-to-use and high-return 3D vision system, forming solutions such as depalletizing, loading and unloading, positioning and assembly, and disorderly grabbing, aiming to provide technical empowerment for global industrial manufacturing and warehousing and logistics automation.
With the continuous upgrading and development of intelligent manufacturing and intelligent logistics, a large number of machine vision needs have been generated, and the penetration of machine vision technology in the manufacturing industry and warehousing and logistics industry represented by the automotive industry has accelerated. At the same time, the increasing complexity of industrial manufacturing and warehousing and logistics application scenarios requires higher and higher performance requirements for machine vision technology, such as recognition accuracy, processing speed, and stability. Machine vision technology is gaining traction, and many machine vision startups are beginning to emerge.
Founded in 2017, Beijing Migration Technology Co., Ltd. (hereinafter referred to as "Migration Technology") has gradually developed into an industry-leading supplier of 3D industrial cameras and 3D vision robotic arm guidance systems with stable and easy-to-use products and fast delivery capabilities by virtue of its deep technology in the field of 3D camera hardware, algorithms and software. At present, the integrated 3D vision solution of software and hardware of Migration Technology has been successfully applied by many Fortune 500 companies in many industries such as new energy, automobile, chemical, home appliances, and metal manufacturing. The growth of migration technology is also optimistic about capital, and after receiving tens of millions of yuan in Series B financing in May this year, hundreds of millions of yuan in financing have been completed.
Recently, the reporter of this magazine walked into the Beijing headquarters of Migration Technology and had an in-depth exchange with Li Hongkun, co-founder and CTO of Beijing Migration Technology Co., Ltd., and asked him to share in depth the development of machine vision technology, as well as the typical application and future development trend of machine vision technology in the field of industrial manufacturing, warehousing and logistics.
Li Hongkun
Co-founder and CTO of Beijing Migration Technology Co., Ltd
Reporter: First of all, please introduce the machine vision technology and its development process, what is the current market demand?
Li Hongkun: From the perspective of technological development, machine vision can be traced back to the 50s of the 20th century, when it mainly focused on the analysis and recognition of two-dimensional images. With the advancement of computer hardware and image processing algorithms, machine vision technology was widely used in the 80s and 90s of the 20th century, and began to be applied in medical image analysis, security monitoring, traffic recognition and other fields. In recent years, with the rise of deep learning technology, machine vision technology, as a branch of artificial intelligence (AI), has ushered in new breakthroughs. Deep learning models, especially convolutional neural networks (CNNs), have achieved remarkable results in tasks such as image classification, object detection, and image segmentation, and have promoted the development of machine vision technology.
From the perspective of market demand, China's manufacturing industry, especially the high-end automation industry, generally has an urgent need to improve the industrial yield rate and "machines help people". The injection of AI will redefine industrial machine vision applications. Although the 2D vision track has shown a trend of giants, as precision manufacturing industries such as consumer electronics, smart cars, and semiconductors continue to improve their accuracy requirements, the defects of 2D vision technology will become more and more obvious.
In general, the current market demand for machine vision technology is rising, and with the complexity of application scenarios, the market has higher and higher performance requirements for machine vision technology, including recognition accuracy, processing speed, stability, etc.
Reporter: What is the background of the establishment of migration technology? What are the characteristics and positioning? What products and solutions are mainly provided for the current market demand?
Hongkun Li: Transfer originates from "Transfer Learning", which is a large class of learning methods in machine learning. In layman's terms, it allows the model to apply what it has already learned from one or more source tasks to a new or related target task. Therefore, we followed this line of thought and began to think about how to combine AI with robots. Through in-depth research, we see that many markets represented by the automotive industry have matured, and there are a large number of scenarios where industrial robots are directly landed; On the other hand, our start-up team began to study 3D vision-related topics at the graduate level, and has a certain amount of theoretical knowledge and technology accumulation, so we finally chose the track of "3D machine vision + guided robot".
In 2017, "Migration Technology" was officially established, focusing on 3D camera hardware, algorithms and software, and is committed to creating stable, easy-to-use, and high-return 3D industrial cameras and 3D vision robotic arm guidance systems, providing technical empowerment for global industrial manufacturing and warehousing and logistics automation.
Although with the rapid development of machine vision technology, there are more and more machine vision suppliers, I think migration technology has its own characteristics: first, it has a self-developed high-performance industrial-grade 3D camera, which can collect high-quality 3D point cloud data; Second, there is mature application software, which can quickly respond to the needs of various complex scenarios, or the needs of different projects in the same scene; Third, it has opened up the entire technology chain, which can achieve stable delivery and continuous supply.
Since its inception, Migration Technology has always positioned itself as a "product company", that is, to deliver standardized products. The advantage is that by packaging documents, hardware, software, etc. into a complete, standardized product, it can achieve fast delivery, help reduce costs in all aspects, and at the same time better establish a good ecological relationship with external partners.
Migration technology family diagram
Under such positioning, the Epic Eye series cameras of Migration Technology adopt completely independent research and development of software and hardware underlying technology, which is suitable for all kinds of robot guidance needs, and has obtained CE, FCC, KC, VCCI and other international certifications. According to the needs of different scenarios, solutions such as depalletizing, loading and unloading, positioning and assembly, and disorderly grabbing have been formed. Among them, the depalletizing solution is mainly used in the whole package unstacking and warehousing, production line transfer, finished product destacking and delivery, and parcel sorting; The loading and unloading solution is mainly used in the machining production line for unloading and unloading processing, process connection and handling; Positioning and assembly solutions are mainly used in production line positioning and assembly, process connection and handling; The disordered gripping solution is mainly used for the sorting of disordered workpieces.
At present, Migration Technology has served many industries such as new energy, automobile, chemical, home appliances, metal manufacturing, etc., and its customers include Mercedes-Benz, Tesla, Volkswagen, CATL, CNOOC, Sinopec, Midea, Gree and many other Fortune 500 companies.
Reporter: Could you please talk about the main application scenarios of machine vision technology? What is the important value for the intelligent transformation and upgrading of the industry?
Li Hongkun: The application scenarios of machine vision technology are very wide. For example, in the field of industrial production, it can be applied to disorderly sorting, loading and unloading, assembly, defect detection, etc.; In the field of logistics, it can be applied to destacking/palletizing, parcel sorting, volume measurement, counting, inventory, etc.; and applications such as 2D vision OCR inspection that are not covered by migration technology. In the transformation of industry automation and intelligence, 3D vision technology, represented by migration technology, has shown great value and potential.
Migration Technology: 3D Vision-Guided Air Conditioning Compressor Handling
Taking the application of 3D vision in the loading and unloading link as an example: there are three main ways of traditional loading and unloading, one is pure manual loading and unloading; The second is the loading and unloading based on 2D vision, which is usually combined with conveyor belts to achieve plane recognition and grabbing, but the disadvantage is that it is impossible or difficult to accurately measure depth information; The third is to realize automatic loading and unloading through non-standard machinery (such as vibrating plate and other special machines), which is not flexible enough. The 3D vision-guided manipulator not only has high imaging quality (especially the imaging accuracy of black reflectors), high grasping accuracy (more than 99.99%), and fast recognition speed (in a certain project, we have shortened the time interval between "taking pictures, recognition, grabbing, and taking pictures" of a black reflective object from more than 10 seconds to less than 7.5 seconds), and the degree of flexibility is very high, which can meet the identification needs of different types of parts and pallets.
Taking the application of 3D vision in the process of depalletizing as an example, the traditional manual operation is not only inefficient, but also has errors and losses caused by human factors, and with the problem of "difficult employment", the labor cost further increases. 3D vision technology, on the other hand, is highly adaptable and flexible. Whether it is cartons of different shapes and sizes, or turnover boxes, drums, metal ingots, bricks, tires and other objects of different materials, the vision system can quickly adapt and accurately identify, provide accurate navigation and positioning data for the robot, so that it can independently identify and accurately grasp objects, and support a variety of stacking methods to achieve efficient depalletizing operations. Depalletizing operations are no longer limited to specific object types or environmental conditions, greatly increasing the flexibility and scalability of logistics handling.
Reporter: Combined with the application cases of 3D vision in the field of warehousing and logistics, please talk about the pain points that the products and solutions of migration technology have solved for customers?
Li Hongkun: Taking the 3D vision-guided depalletizing solution provided by Migration Technology for an integrator in the refractory brick material industry as an example, the traditional depalletizing method is often limited by problems such as low efficiency and poor safety, so the customer's 3D vision technology mainly focuses on high efficiency, safety, meeting the production rhythm, and stable and reliable quality. This project uses 3D vision-guided robots to accurately identify, locate and grasp refractory bricks, so as to achieve automated, efficient and intelligent destacking operations.
1. High compatibility and flexibility
The 3D vision depalletizing solution of Migration Technology can be compatible with a variety of types of refractory bricks. With the help of intelligent algorithms, the system is able to automatically identify and locate up to 99 different brick types, while accurately identifying the brick orientation. This feature allows the solution to meet different production needs without frequent equipment changes or complex adjustments, which significantly improves the flexibility and adaptability of the production line.
2. Achieve 4 seconds of recognition and positioning, with an accuracy rate of 99.9%
Migration Technology Vision Systems have excellent processing speed and accuracy. The rapid identification and positioning of refractory bricks can be completed in about 4 seconds, and the accuracy rate of recognition and guidance is as high as 99.9%, which greatly reduces the error rate and rework rate, thereby improving the overall production efficiency.
3. Efficiently adapt to the rhythm of the production line
The efficient vision system of Migration Technology perfectly adapts to the rhythm of the production line, and realizes a fast grasp of the beat every 8 seconds. The efficient operation mode not only increases the throughput of the production line, but also ensures the stability and consistency of the depalletizing process.
4. Flexibly switch the grabbing mode
The migration technology depalletizing solution can flexibly switch the grabbing mode according to the needs of the production line. Whether gripping piece by piece or group by set, the system guides the robot arm to precise gripping and placing to ensure maximum efficiency in every operation.
Up to 6 bricks can be grabbed at the same time in a single operation, which is particularly important when processing large quantities of refractory bricks efficiently, further increasing the productivity and flexibility of the line.
5. Easy to interact and easy to use
The vision software of Migration Technology has excellent interactivity and ease of use, and it is easy to replace the model of refractory bricks, which helps front-line workers to operate and use it easily, which not only reduces the difficulty of operation, but also reduces training time and cost.
Reporter: Combined with the overall application of machine vision in the current industry, what are the characteristics and development trends of machine vision technology? What are the main upgrade directions?
Li Hongkun: From the perspective of the workflow of the machine vision system, the camera is first required to shoot a point cloud image with high enough accuracy and complete enough in the shortest possible time; Secondly, the visual algorithm is required to process the image to identify the specific position and posture of the object in space. When a plurality of objects are recognized, the grasping planning of the objects is further involved, such as grasping objects, sequences, etc.; In addition, it also involves the planning of the robot itself, such as whether the robot will collide with environmental objects in the process of performing actions, and the specific operation path of the robot.
Therefore, machine vision mainly shows the characteristics of high precision, high efficiency and high stability. Among them, "high precision" is the key to ensure the accuracy of tasks such as item identification, positioning, and measurement in the logistics process; "High efficiency" is an inevitable requirement for the efficient and high-paced operation of the logistics industry; "High stability" is an important guarantee to ensure the stable operation of the machine vision system for a long time and reduce the failure and maintenance costs. In addition, machine vision also needs to have strong adaptive capabilities to cope with recognition tasks in different environments and under different lighting conditions.
In terms of development trends, with the continuous development of deep learning, artificial intelligence and other technologies, the application of machine vision will be more intelligent and autonomous. On the one hand, through deep learning and big data analysis, the machine vision system can continuously optimize the algorithm to improve the recognition accuracy and speed. On the other hand, machine vision will be combined with robots, automation equipment and other technologies to achieve more efficient and accurate logistics operations. At the same time, with the popularization of Internet of Things technology, machine vision will be more closely integrated with other logistics information systems, promoting the digital transformation of the logistics industry.
In this process, Migration Technology will always focus on what we are good at, do a good job as a 3D vision system supplier, and work with partners to establish a good image for the 3D vision industry in the minds of customers, so as to achieve a win-win situation for customers, partners and ourselves.
Reporter: How do you see the future application prospects of machine vision technology? In what dimensions will migration technology make further efforts to create more value for the industry and customers?
Li Hongkun: I believe that machine vision technology, especially 3D vision, will have broad development prospects.
First of all, the scenario-based demand is much higher than the industry demand, and the scenario-based demand is not divided into industries, and the solutions are very similar, such as the destacking scenario, the logistics industry has the scenario of unpacking cartons, and it can also be used in the chemical industry to unpack the packaging bag and the food industry to unpack the sack...... These scenarios can be used in almost the same way, but the difference in camera model and mounting height is different.
Secondly, the "employment difficulty" problem encountered in the logistics industry today has forced many small and medium-sized enterprises that did not dare to carry out automation transformation before to turn to automation, which also provides a certain market opportunity for the Party B market.
Finally, according to Interact Analysis's latest Machine Vision research report, 3D cameras are expected to lead the growth of the global machine vision market in the next five years. In 2022~2028, the compound annual growth rate (CAGR) of the global 3D machine vision camera market is expected to be 13%, much higher than the expected growth rate of 6.4% in the global machine vision market. Global sales of 3D cameras are expected to grow from $767 million in 2022 to nearly $1.6 billion by 2028.
After years of deep cultivation and accumulation, Migration Technology has formed a strong competitive advantage in the imaging quality, imaging speed, recognition algorithm, software design and other aspects of the 3D vision system, giving automation equipment a pair of "eyes" that can still see more clearly and three-dimensionally even in the dark. In the future, we will continue to make efforts in these aspects: first, the upgrade of hardware equipment, including cameras with higher resolution and larger field of view, higher resolution, larger field of view, and smaller weight and volume; the second is the optimization of algorithms, which improves the accuracy and speed of recognition through advanced technologies such as deep learning; The third is the upgrade of the software system, including a more user-friendly interface and more flexible configuration options.
In short, in addition to ensuring product stability and reliability, ensuring a high project delivery rate, and always paying attention to technology research and development, we will further help integrators and agents to complete market co-construction in various aspects, such as migration technology to undertake innovative application exploration in new industries and new scenarios, and assist integrators to complete the construction and delivery of new business scenarios. In the future, Migration Technology will also shift from the current "passive going overseas" to "actively going overseas" to serve industrially developed countries such as Japan, South Korea, and Eastern Europe, as well as emerging markets such as Southeast Asia and North America, so as to create value for global customers.