Author | Xu Zhi Editor | Liu Wei
At the Industrial Automation Exhibition of the China International Industry Fair 2026, Hikvision's booth was surrounded by crowds. Over 100 scenario digitalization products were displayed before the audience, ranging from the sub-micron precision semiconductor packaging 3D AXI inspection equipment, to the high-definition visual pyrometer capable of clearly seeing the industrial site under the extreme high temperature of 3200℃, from the auscultation optical fiber that "diagnoses the pulse" for belt idlers, to the scanning imaging radar that performs 3D modeling for silos.
Some people sighed, "Hikvision is really no longer just about security," while others stopped in front of this "highly AI-infused" booth, trying to figure out a question: What gives this company, which established itself in the security field, the right to take such deep roots in the industrial sector? To answer this question, it is better to shift our focus from the booth and look at the changes happening at industrial manufacturing sites.
Scenario Digitalization: A Mandatory Question for Industrial Upgrading
Over the past two decades, Internet digitalization has profoundly reshaped industries such as consumption, social interaction, finance, and entertainment. The online transformation of people and activity information has given birth to the world's highest market cap technology companies and changed the daily lives of almost everyone. But this wave has a boundary; it has completed the digitalization of people and activity information. The physical world itself—those massive factories, operating production lines, high-temperature furnaces, and belts stretching for kilometers—has a long way to go in its digitalization process.
Currently, AI large models and multi-dimensional perception technologies are accelerating towards maturity, and the digitalization of IoT data in the physical world is releasing huge industrial space. According to CCID Consulting, the market size of China's Industrial Internet reached CNY 1,258.78 billion in 2025, a year-on-year growth of 12.6%. It is expected to reach CNY 1,675.35 billion by 2028, and the Industrial Internet will further evolve towards intelligence, data-driven, and system-level collaboration.
This is a sector with a long runway and immense potential.
Signals at the policy level are equally clear. The outline of the "15th Five-Year Plan" proposes "implementing large-scale application demonstration actions for new technologies, new products, and new scenarios, and increasing the cultivation and opening of scenarios." The 2026 "Government Work Report" further proposes "building a new form of intelligent economy," emphasizing the deepening and expansion of "AI +." The value of scenarios lies precisely here. It moves technological innovation from the laboratory to real scenarios, becoming a "testing ground" for new technologies and new products, building a bridge between R&D and the market, and accelerating the transformation of scientific and technological innovation achievements into real productive forces. The upgrading of China's industry from "comprehensive" to "strong" precisely requires this kind of scenario thinking.
China has 41 major industrial categories, 207 medium categories, and 666 small categories, making it the only country in the world that possesses all industrial categories listed in the United Nations industrial classification. This characteristic of being "comprehensive, diverse, and large" means that the fragmentation of industrial scenarios far exceeds imagination. The era when a single universal solution could conquer the world is over; technology must be tested in real scenarios and take effect in specific problems. This is precisely the underlying logic behind Hikvision's advocacy of "scenario digitalization." Scenarios are the foothold, and digitalization is the path.
Many manufacturers in the industry talk about "scenarios" or "digitalization" separately, but it is rare to combine the two to form a unified industrial cognition. The reason Hikvision can clearly put forward this concept is that it simultaneously possesses several conditions: starting from visible light and expanding to a multi-dimensional perception technology system including infrared, X-ray, millimeter-wave, and acoustic waves, it can penetrate the physical world to help users collect various data; the Hikvision Guanlan large model has achieved deep integration of multi-modal large models and industry knowledge, enabling it to read data and solve problems in specific scenarios; more importantly, the technology extension logic means that almost all of Hikvision's industrial technologies can find mature technological predecessors in the security field. Security needs to "see anomalies clearly," while industry needs to "perceive data"; security needs "intelligent early warning," while industry needs "predictive maintenance." Both require the combination of perception and intelligence.
The "Flywheel Effect" of Perception and AI
After understanding the importance of scenario digitalization, the next question is: What gives Hikvision the right to do this well? The answer lies in its perception and AI capabilities accumulated over many years.
First lies the comprehensiveness of perception technology. The fragmentation of industrial scenarios means that a single perception method is often difficult to adapt to complex working conditions. Just as humans cannot understand the world solely by their eyes, information acquisition in the industrial field also requires the coordination of "eyes, ears, nose, tongue, and body." Hikvision has built an IoT perception technology system covering multiple fields such as visible light, infrared, X-ray, millimeter-wave, fused spectrum, and acoustic waves. The company has over 30,000 models of hardware products and more than 100 software products on sale, which can meet the differentiated needs of different scenarios in various industries. The value of this multi-dimensional perception capability is reflected in the chip quality inspection scenario.
Aiming at the typical complex scenario of chip inspection, Hikvision has built a complete product system from macro to micro, from internal to surface, and from measurement to material analysis: the high-precision semiconductor packaging 3D AXI inspection equipment has sub-micron level transmission inspection capability, which can accurately identify deep chip defects such as internal gold wires, solder joints, and penetrating defects; the ultrasonic microscope can identify micron-level defects and has extremely high sensitivity to planar layer crack detection; the super-depth-of-field 3D digital microscope can clearly capture micro-defects such as scratches and glue overflow on the chip surface; mass spectrometry and chromatography equipment can accurately detect whether organic pollutants such as photoresist and flux remain on the surface of chip pads...
Beyond perception, AI large models have endowed Hikvision with the ability to solve "chronic and difficult" problems. The core dilemma faced by traditional industrial inspection solutions is: each scenario requires separate data collection, annotation training, debugging, and deployment, with cycles lasting up to several weeks and high costs, and product changeovers require starting from scratch. The few-shot learning and strong generalization capabilities of the Hikvision Guanlan large model have significantly lowered this threshold.
Taking Hikvision Ruiying's X-ray industrial inspection large model as an example, in the past, inspecting a single object required thousands to tens of thousands of annotated images; now, only dozens of samples are needed to complete model training and deployment. In some special scenarios, relying on unsupervised large model technology, only 3 to 5 samples are needed for rapid online deployment, greatly improving iteration efficiency. The large model is pre-trained based on massive multi-industry data, mastering the general essential characteristics of defects. It can distinguish between "real defects" and "normal forms that look like defects," with clearer decision boundaries and recognition logic more aligned with industrial real scenarios, effectively reducing false alarms and improving overall inspection accuracy.
Similarly, in the field of ultrasonic inspection, Hikvision has independently developed a knowledge-enhanced signal reconstruction network and an ultrasonic large model. Through methods such as adaptive noise separation, prior knowledge embedding, self-supervised pre-training combined with few-shot transfer, and automatic quantitative analysis, AI can accurately read waveforms and signals like a senior master. In the NEV parts inspection scenario, the inspection efficiency of a single workpiece has increased by more than 60%, and the defect recognition accuracy is not less than 99.5%. This superposition of "perception + AI" capabilities has formed a unique positioning for Hikvision in the industrial field: it can not only "see" various elements in industrial scenarios using multi-dimensional perception methods but also "understand" the meaning behind the data using AI large models.
These two core technologies, IoT perception and AI, are like a pair of "flywheels," driving each other. The development of IoT perception technology digitalizes the physical world, providing scenario support for the leap in AI capabilities; the development of AI capabilities can continuously improve the performance of IoT perception devices and give birth to a large number of industrial-grade applications.
Answers in Scenarios
Whether capabilities are effective must ultimately be tested in scenarios.
In the field of quality inspection, the hinge inspection of foldable screen mobile phones is a landmark scenario conquered by Hikvision Ruiying. At the end of 2022, foldable screen hinges were still an emerging process. The product structure was precise and complex, defect characteristics were subtle, and imaging interference was extremely strong. The production process derived more than ten types of difficult defects, which could not be effectively identified by the naked eye and traditional optical inspection equipment. Hikvision Ruiying formed a special task force with customers. Based on the X-ray large model intelligent inspection system, it achieved stable detection of various defects such as cracks, pores, and foreign objects, with a minimum accuracy of 0.05mm. Through X-ray AI full inspection, the defect rate dropped by an order of magnitude, significantly improving product yield. After the solution was first implemented in leading benchmark customers, it rapidly expanded to the entire mobile phone industry.
In the field of production operations, the production line behavior SOP inspection solution has already gained momentum for scaled replication. Under the trend of flexible manufacturing, the same production line quickly switches categories and schedules production on demand. The operation standards for different products, workstations, and materials vary greatly, making how to ensure product quality a difficult problem.
The SOP behavior inspection system solution created by Hikvision comes with out-of-the-box common production line algorithms. Facing new workpieces and new scenarios, it only requires 1 to 8 images for adaptation, achieving minute-level deployment. After being launched in more than 600 core workstations at Hikvision's Chongqing, Tonglu, and Wuhan production bases, the efficiency per workstation has increased by up to 80% compared to the original manual method. Jifeng Seats in Li Auto's supply chain system has significantly reduced the missed driving rate of C-type nails for seats to a minimum of 0.01% through the SOP behavior inspection system, and has reused the solution for upstream and downstream supplier teams. Power system supplier United Power has covered 100% of key operation processes, from stator assembly to terminal crimping, with AI conducting full-process inspection and real-time verification of operation specifications. Currently, Hikvision's SOP solution has cumulatively landed hundreds of projects nationwide, covering multiple industries such as automotive parts, 3C electronics, semiconductors, and food packaging.
In the field of equipment monitoring, Hikvision's industry-first positioning auscultation optical fiber provides a "deep water zone" solution. There are about 300,000 industrial belts in China, and some long belts cost over CNY 100 million. Traditional inspections rely on human eyes and ears, with harsh environments, high noise, and heavy dust. Long belts stretch for several kilometers, making inspections time-consuming and difficult. Hikvision innovatively installed optical fiber sensing probes on belt conveyors, accurately locating faults by collecting idler sounds combined with AI algorithms. This technology won the only gold award at the 2025 National Innovation Method Competition and has been implemented in multiple enterprises such as Nanjing Iron and Steel, Huaxin Cement, and Huadong Materials.
In the field of robotics and radar, Hikvision's market validation data is equally solid. In 2025, Hikrobot's annual revenue reached CNY 6.452 billion. As of now, Hikrobot's vision products have cumulatively shipped over 10 million units, and mobile robots have exceeded 200,000 units off the production line. In the domestic market, one out of every two industrial cameras and one out of every three mobile robots comes from Hikrobot, cumulatively serving over 20,000 customers. Hikvision's industrial radar extends from technical accumulation in the vehicle field, building a complete product system from 1D single-point ranging level gauges to 2D cross-sectional imaging radar, and then to 3D scanning imaging radar and 4D obstacle avoidance radar.
At the Shangluo Power Plant in Shaanxi, the laser scanning imaging radar can complete a comprehensive inventory check in just 3 minutes, with an accuracy rate of over 97%, improving the coal inventory efficiency by dozens of times compared to the past. It is worth mentioning that Hikvision itself is also the most direct testing ground for these technologies. Hikvision's smart factory has over 30,000 models of hardware products, processing an average of 14,000 orders daily. Each order averages only about 30 units, and the proportion of customized configurations is as high as 70%.
Facing this production characteristic of "multi-variety, small batch, and large-scale customization," Hikvision has deeply integrated its self-developed IoT perception, AI large models, and other technologies into the entire manufacturing process. Over 1,500 mobile robots work collaboratively, and the fully automated unmanned production line with smart cameras has improved efficiency by 243% compared to traditional lines. In 2025, Hikvision's smart manufacturing factory was selected for the first batch of national pilot-level smart factory cultivation list.
Conclusion
The previous wave of Internet digitalization completed the online transformation of people and activity information, releasing huge industrial dividends. The digitalization of the physical world is opening up an equally broad industrial space.
Every piece of equipment, every production line, and every process in the factory contains the demand waiting to be perceived, understood, and optimized. What Hikvision wants to do is to build that bridge between the physical world and the digital world.
Using multi-dimensional perception capabilities covering visible light, infrared, X-ray, millimeter-wave, and acoustic waves, it transforms information from the physical world into analyzable digital signals; using the Hikvision Guanlan large model to read the meaning behind these signals, assisting decision-making and guiding execution, enabling industrial manufacturing to truly achieve the leap from "seeing" to "understanding." The digital and intelligent upgrading of industry begins with every real scenario.
From the over 100 products on the booth, to the workstations on the production lines, and then to the segmented application scenarios in thousands of industries, Hikvision's scenario digitalization practice is answering a fundamental proposition: Only when technology steps out of the laboratory, enters the industrial site, and goes into every specific production link, can the value it creates truly take root. And this is precisely the most solid foundation in the digitalization wave of the physical world.