Video Surveillance: What Will China Do Next?
Top image - China's explosive mobile internet revolution is the real engine behind the country's push into AI and big data. That revolution started just 5 years ago...
IF you’re wondering why you didn’t get SEN EDM last Friday, it’s because we spent last week in China visiting major CCTV and electronic security companies seeking a sense of where the market is heading over the next 3 years. We’ve got a lot of material for detailed features in the future but these are our initial impressions.
We started in Hangzhou with Hikvision, Dahua and Uniview, then flew to Shenzhen to visit TVT and Sunell. The plan was to see the latest products, speak to the companies about their future directions and take a trip to each factory to get an idea of scale and the latest manufacturing processes. We spent time with the teams of each company to discuss wider matters. It was an interesting experience – I quickly realised the notion of Chinese manufacturers as a composite block is erroneous. All these companies are quite different.
Sitting on the ‘plane flying out of Hong Kong attempting to synthesise the meaning of what we saw during the week, I found it impossible to limit my ideas about what’s going on in China’s CCTV market to mere security electronics. Instead, I ended up thinking about broader technological trends – some global but many unique to China’s explosive smartphone-driven internet revolution. There are cultural, historical, business and political themes that feed into what’s happening there. China’s huge domestic market, phenomenal industrialisation, preparedness to copy and develop good designs, combines with an intensely competitive local business environment where any supplier that survives will be a global powerhouse.
I visited Hikvision back in 2014 – 5 years is an eon in modern electronics and computing – it’s worth pointing out the scale of change that has occurred in that time. In 2014 Hikvision’s manufacturing still had a considerable element of manual labour in it – there were SMT lines, but the finished components were put together by hand. That’s changing fast. Hikvision’s factory features a huge number of automated production lines – there are still techs on the automated lines doing jobs that require too much dexterity for robots or resolving problems when they arise, but even IP66 ratings tests are now automated.
I stood in the corner of the company’s automated workshop, with its low ceilings and heavy pillars and looked at row upon row of automated lines stretching into the distance – I started to lose count in the mid-20s. These automated lines are built by Hikvision’s own robotics division. The scale of the facilities is immense.
The Hikvision team told us their future is AI – we heard that a lot on this trip. All the Chinese manufacturers are focused on AI – the camera as a sensor – with clear operational outcomes tailored for security and vertical applications such as retail. There’s face recognition, line crossing, object searching, perimeter protection and all the rest. None of this is ground-breaking – global competitors have been developing these functions, too – what’s new is that AI is finally reliable and affordable. This AI is locally developed and is not limited to high end cameras – it’s being pushed into every layer of the product stack. Although there are specialised servers being built to handle large scale AI, most of what we saw was camera-based.
While AI is a big thing in China it’s not AI in the way most people think. Since 2016, there’s been a profound push towards enhancing efficiencies and liberating functionalities using AI – it’s become a national endeavour. We’re not only talking about face recognition here. There’s been huge investment in AI laterally and with the development of the latest ‘narrow’ neural AI algorithms – deep learning – AI is becoming more and more capable in security applications. But AI is not just in the products – it’s leveraged in making the products on automated lines and transporting finished stock around sites on automated vehicles.
Hikvision also pointed out that cloud and big data would be a major part of its plans over the next 3 years. Both these focuses are very interesting on many levels. Chinese internet companies have been prepared to do things for free in a way that could revolutionise the cloud business model globally. Think of affordable security, automation and CCTV with a free cloud back end and a freemium paid by high volume users. Of course, to do this well, the players will need global penetration of Chinese platforms.
A consideration of big data is that deep learning gets stronger the more data it’s fed and nowhere has more data than China. There are more than 500 million internet users in China, and they use mobile phones to buy off-line services online – food, haircuts, share rides and much, much more. This means that Chinese AI functionalities can be polished endlessly in huge local applications. Apps aren’t simply imposed on the market as we see in Australia but are relatively simple at release and then get improved by analysing the AI-recorded use-patterns of users in real time to constantly add enhancements.
We-Chat is an example of China’s different thinking when it comes to modern technology. It’s an umbrella app that supports messaging, video, photos, ride bookings, bike hire, food purchases, billing, booking dental appointments, purchasing airline or train tickets, news services, and loads more. To make all this run, We-Chat has an integrated digital wallet so We-Chat users never have to leave the app to manage their lives. The data purchases and actions generate feedback into better AI and a more useable app.
China is more comfortable with AI than many other countries. Recently Microsoft’s president Brad Smith said countries should regulate AI such as face recognition before democratic freedoms are impacted. According to Smith, “We are the first people in the history of humanity who will give this power to machines…so, for the future of humanity, it is essential we think this through”. Meanwhile, Yoshua Bengio’s ethical guidelines for AI – embodied in the Montreal Declaration – call for ethical use of AI. Benigo told Nature magazine recently that use of AI in surveillance — which he said could have potentially positive benefits – is also open to abuse.
In comparison, Chinese entrepreneurs are throwing themselves at AI with all their might and using it to resolve operational business problems and uncover profitable niches in every way they can imagine. Given this dichotomous position on AI – San Francisco is considering banning the technology in government applications – it’s hard not to see that the future of AI is going to be more China’s future than anyone else’s – as important as the points Smith and Bengio make may prove to be. The material point here is that resistance to AI outside China will lead to a lack of investment in AI, feeding into reduced functionalities and reduced efficiencies from competitors.
Back on our tour, Dahua’s factory was also revealing – it’s big and we saw a lot more of the production floor. There’s plenty of automation, then there are semi-automated and manual production lines. Again, Dahua’s in-house robotics team is building robotic production lines to its own designs and to meet its own requirements. Dahua, too, employs AGVs to move components and stock around the factory. I got the clear sense China is going to break the developing world’s mould of low-cost labour with its combination of robotics, automation and AI. This combination is also going to raise the bar for competitors – not only overseas but locally. Few will be able to liberate the sort of resources required to reach this level ever again.
And automation is going to drive quality up while driving prices down. It’s a ground-breaking dichotomy the market has never seen before. Dahua is focusing on AI, as well as on vertical solutions and big data. When it comes to the 2 largest manufacturers – Hikvision and Dahua – there’s a temptation to see the lock-step of technological development in a less than positive light. That’s a mistake. There’s an intensity of competition going on here, a cadence of development and counter development. Will such competition one day deliver a camera that can do almost anything? When you look at a phone like Huawei’s CP30, it’s impossible to reject that notion out of hand.
Uniview is smaller than these 2 giants. The company bought a large facility in Hangzhou 12 months ago and is currently bringing together its manufacturing and distribution there. Uniview is also investing heavily in AI, which it has developed in-house. Uniview’s Ximen Yan made the case that the next 5 years would see CCTV technology pushed forward almost solely by the Chinese manufacturers. It wasn’t easy to form a counter argument to these assertions.
The tight production line we saw at Uniview was smaller than the other 2 and there was more hands-on work. But the same processes were all there. Uniview is strong in government applications and its solutions reflect the needs of those clients – Uniview has some powerful solutions, as well as considerable experience in major systems that’s often overlooked outside of China. The company’s attainment of GDPR certification is a key development, too.
TVT is different again – small by Chinese standards but huge compared to international competitors. TVT is also investing heavily in AI and has a comparatively large R&D department. TVT is focusing on verticals, tailoring its solutions to meet the specific needs of a predominantly commercial client base. TVT’s core focus is making the cutting edge of CCTV – H.265, AI – affordable for end users. TVT has also achieved GDPR certification and the company is a leader in software development – it was the first to come out with a smart phone app.
Sunell is the oldest of this group and the smallest. Founded back in 1997, its comparatively compact range has strengths in thermal and hemispheric cameras. Sunell has a comparatively huge R&D department and has developed its own AI algorithms, too. Last year the company shipped tens of thousands of AI-enabled cameras to commercial customers. People in Australia are probably much more familiar with Sunell’s product than they realise – for many years the Pacom OEM range was Sunell. Sunell’s thermal cameras and interface also caught my eye during our visit. The hardware is extremely compact and the resolution and contrast are solid. Sunell is engineering-based and privately-owned, same as TVT, and both businesses are built primarily on commercial work.
All these manufacturers have wide ranges. It’s easy to look at the Chinese CCTV makers’ enormous product spreads – Dahua makes 1000 different products – and think they’re produced by an R&D blunderbus fired without much thought. There’s a certain truth to this wide-angle view of the market – it applies to the test-every-niche nature of Chinese tech startups, too – but there’s much more to it than that. These manufacturers are end user-driven and with a huge local market to please, their product diversification and endless polishing of functionalities is a means of staying ahead of the hungry local competition.
The next 3 years of Chinese CCTV development is not likely to look the same as it did over the last 5 years. The market has sorted itself into a small group of highly competitive manufacturers. In a real way, the last 5 years put a giant broom through the Chinese CCTV market – these 5 companies and perhaps 1 or 2 others (including big telco, Huawei), are the last players standing in an epic struggle for survival in the largest CCTV market on Earth. Dozens of CCTV manufacturers have foundered in a few short years.
In my opinion, we’re going to see an evolution of business models alongside the evolution of camera technology moving forward – that’s going to be fascinating. It’s also going to make the potential functionalities of CCTV wrangled by AI available to every user. You don’t need to talk to many security managers to discover that this is something they all want very much. Just how the Chinese plan to bring these solutions to the rest of the world remains to be seen but its likely to be left field when it arrives.
Something I couldn’t avoid noticing was the work ethic in China – 12-hour days, and 6-day weeks are normal here. Whether this makes workers more or less productive is something for the better informed to debate but these long hours make for faster decisions because the team is nearly always ‘on’. Further, their combined R&D investment is enormous. This is a key aspect – the ability of companies to develop and rapidly bring to market products that combine the latest breakthroughs and to relentlessly polish those offerings to ensure they stay at the cutting edge.
Visiting China’s big CCTV manufacturers was a fascinating experience. Whatever you think about AI used on a wider scale, it’s a powerful tool in smaller applications, too. Security managers around the world are hungry for the efficiencies it brings to their applications. In the future, that hunger is going to be met – and met very affordably indeed.