Surveillance Redefined: Advances in Vision AI Drive China’s Leading Security Market

Rapidly advancing automation technologies like computer vision have enormous potential for dramatic societal impact, and need to be applied with caution. Facial recognition is one of the most scrutinized examples today of applied computer vision. To create a foundation for social and political analysis, we want to take a birds-eye view of the differences in the technology ecosystems of the United States and China and understand how they are evolving.

In only a matter of years, China has seized a sizable commercial market for vision AI as well as the world’s most valuable vision AI startups. This is largely due to the rapid adoption of commercial AI applications in security, surveillance, smartphones, and finance. In particular, vision AI startups first commercialized their technology in security and surveillance applications.

China is the world’s largest manufacturer of video security and surveillance equipment and has extensively incorporated such equipment in public spaces. While some Chinese surveillance equipment manufacturers first started to incorporate AI into their products around a decade ago, advances in deep learning in recent years have significantly improved the accuracy of computer vision and video analytics.

Until around 2010, foreign companies were leading China’s security equipment and solution market. However, the market turned to favor domestic suppliers thereafter, because foreign companies failed to localize R&D and it became extremely difficult for foreign companies to access data collected by public projects. In addition, it was difficult for foreign companies to navigate through Chinese regulations on data collection and usage.

While it was hard for conventional foreign security companies to make a comeback in security specifically, new opportunities emerged for foreign companies in related industries such as AI-powered video storage and chips. In 2016, HikVision introduced the processor chips for computer vision developed by Caniforlia-based Movidius, which was later acquired by Intel. Qualcomm also invested in SenseTime and began collaboration with the company on on-device AI.

Popular security and surveillance companies such as HikVision, the world’s biggest surveillance camera maker, and Dahua Technology have become major players in AI with an abundance of new equipment and solutions. HikVision claims that their AI-powered products have been used ubiquitously throughout China and announced earlier this year that they would soon open up their AI platform.

Some other major players in telecom and electronics equipment sectors who have invested heavily in AI, like telecom giant Huawei, are now competing with incumbents in security equipment.

Largely due to the commercialization of vision AI technology in security and surveillance, several vision AI startup unicorns have emerged in the last few years, including SenseTime, YITU Technology, CloudWalk, and Megvii. SenseTime is now the world’s most valuable AI startup. Its technology has also attracted aforementioned players in the security sector for several joint projects. YITU also launched an AI security platform in conjunction with Huawei in earlier this year.

Vision AI-powered security capabilities are also important part of the ambitious smart city projects being developed by established tech giants like Alibaba, Huawei and Baidu. These projects will largely depend on vision-based IoT devices for constant data flows and analytics capabilities that can be enhanced by AI.

The public services sector is currently a major customer for AI-powered security solutions, which have been adopted in public safety and surveillance programs, law enforcement, transportation, financial fraud detection, and more. CloudWalk claims that its facial-recognition-enabled surveillance cameras have been implemented in over 80% of domestic airports. Over 30 provincial and 150 municipal public security departments have adopted the technology provided by YITU to identify thefts and hunt fugitives. The AI-powered surveillance systems are increasingly used by cities to find missing children or seniors and monitor and manage traffic. There are also more controversial projects hosted by the government like the surveillance of Xinjiang, a province heavily populated by muslims, and the chase of BBC reporter John Sudworth.

Rapid adoption of vision AI was driven by the technology’s improved accuracy. SenseFace, the facial recognition solution developed by SenseTime, which has been adopted by many local public security agencies claims, has increased its accuracy rate by 30 times. Qingtingyan, developed by YITU, achieved its first deal with a local public security agency after the company increased its accuracy rate in identifying cars with swapped plates by 60%.

Through these projects, startups, and established tech companies, China has accumulated access to vast amounts of data collected by the public sector, big tech companies, and existing surveillance equipment makers which is critical for AI development. For example, YITU claims that it has collected over one billion facial images.

Chinese authorities have issued documents to promote AI and other related technologies in public security since 2015. A 2015 decree required the surveillance camera coverage rate in critical public places to reach 100% by 2020.

Just last year, only 30% of Beijing was covered by surveillance footage. Now, 100% of the country’s capital is monitored by 46,000 surveillance cameras.

The growing private security market — particularly in corporate security and smart home sectors — is proliferating the growth of China’s security market. Driven by AI advancements, China’s security market is projected to increase by over 10% annually to reach about RMB 1 trillion (US$150 billion) in 2020.

The Comet Labs Research Team in Beijing explores industry and AI technology trends in China, and seeks to understand how they might serve as helpful cases for startups in the United States and around the world.

If this got you thinking and you’d like more, hit the follow buttons below and subscribe to the Comet Labs newsletter for a bi-weekly digest of our favorite AI and robotics stories.