AI temperature detector

As the COVID-19 pandemic becomes a major public health event, security companies are joining the fight.

It is also a training ground for security and AI technology. On the timeline of racing against time, enterprises quickly develop algorithms, design products, integrate supply chains and finally deploy to the ground. After this battle, the intelligent security industry will accelerate differentiation.

The epidemic is at the moment, but also the Spring Festival holiday, security enterprise team “upstream”, in the coordination of the government, AI temperature measurement project effective scale: quickly complete the algorithm research and development, solve the supply chain problems, to achieve the landing verification…… A field of emergency AI war epidemic story staged one after another.

However, in this big test, the shortage of supply chain and the start date have been postponed again and again, bringing no small challenge to the actual implementation of THE AI thermometer. On the one hand, the urgent need for public health facilities, on the other hand, the numerous objective restrictions of the epidemic situation, how is the actual implementation of the AI thermometer? What new opportunities will it bring to the security industry in 2020?

Emergency needs in new scenarios

If you’re already on your way home from train stations, airports and subway stations, you’ll notice that there’s a new camera at the security checkpoint.

Yes, it’s an AI thermometer, with a tripod, a camera and a laptop cobbled together in its basic form.

When each pedestrian passes by, his temperature will be displayed on the display screen. When the detection temperature is normal, no sense can be achieved. Once a pedestrian suspected of high temperature is detected, the system will automatically alarm, and then the security personnel will conduct a second temperature measurement.

In order to effectively cope with the epidemic prevention and temperature measurement requirements of high human flow, THE AI temperature measurement project has the two-way support of infrared and artificial intelligence technology: infrared technology ensures the demand of temperature measurement at a distance without contact, and computer vision technology realizes fast positioning by accurately identifying the forehead of personnel.

How to ensure both efficient passage and accurate temperature measurement is the proposition of the new generation of “security +AI” enterprises tackling the epidemic.

In baidu, including 29 start AI temperature measuring r&d, multiple departments hundred people fight, evening night finished for the day wore monitoring model of the outbreak, completed within three days from the video data samples to the testing model building and so on basic research and development work, followed by annual 4 in Beijing qinghe pilot deployment, the railway station to meet Beijing travel peak.

Kuang Shi, CEO Yin Qi personally took charge of the project and set up a working group of more than 100 people to complete the preliminary design of AI temperature measurement on the eve of Chinese New Year. The algorithm development and internal debugging of the screening system were completed before the third day of Chinese New Year. After more than 10 days of overnight efforts, the whole system was officially launched on February 4.

In the fight together with deep pupils, the second Beijing Science and technology Commission of an emergency r & D phone, and THE IDEA of CEO Zhao Yong happened to coincide, more than 30 people to establish the team quickly, just 8 days to develop AI temperature measurement system, the product has been in Beijing West station “mount guard”.

According to the heart of the machine, in order to rush the AI temperature measurement project, many employees directly eat and live in the company, some sleep only 2~3 hours a day… At present, no matter Baidu, Sensetime, Megvii, Gling Deep Tong, or Autonavi Infrared, Hikvision, Dahua, Yushi and other companies, have launched AI infrared temperature measurement products.

From the market solution, the industry mostly uses infrared/visible dual sensors, combined with infrared thermal imaging and face recognition, in a non-contact way to achieve multi-target temperature screening, temperature measurement accuracy is mostly within ±0.3℃.

From the technical logic, it needs AI algorithm to identify and track the human image, that is, accurately locate each face in the camera picture, and then calculate the temperature of the face (mostly forehead) combined with the infrared temperature dot matrix.

The details involved in face recognition under masks, ReID (pedestrian re-recognition), and the calibration between the two cameras directly affect the effect of each scheme.

From the industry’s positioning of AI temperature measurement products, the value it plays is still the initial screening, for suspected high temperature personnel, still need to manual second temperature verification.

The reason why it is positioned as preliminary screening, on the one hand, preliminary screening + reinspection can enhance security from the process; On the other hand, due to the limitation of product precision and complex environment, the product availability is also affected to a certain extent.

In crowded scenes such as airports and railway stations, once the crowd is out of the view of the camera, temperature detection cannot be carried out. Therefore, during the deployment of AI temperature measurement products, personnel guidance is also required, or entrances and exits are set through isolation belts to ensure that temperature measurement is not omitted.

Of course, there are still some limitations to the quick-start screening scheme.

The differences in technology and engineering bring about temperature errors at different levels. Although the errors of most schemes are ±0.3℃, the errors of some products are still up to ±0.4℃ and ±0.5℃, which will have a certain impact on availability. In addition, the operating temperature of the infrared sensor is usually between 0 and 50℃, which will affect the accuracy of temperature measurement if the temperature is too high or too low. Therefore, outdoors may not be the ideal landing scene for AI temperature measurement equipment in northern areas in winter.

Although cannot replace artificial, but AI temperature products value is still important. In the scene of high traffic, the close security inspection itself has great risks, and the risk of crowd squeezing will be greater. Preliminary screening of body temperature can also reduce the work pressure of security personnel.

Chen Tianbo, product director of Gling Deep Pupil, also said that through the company’s tests, the AI temperature measuring system has the same effect as the forehead temperature gun. In his opinion, a good AI infrared temperature measurement scheme should meet at least three conditions:

First, the temperature measurement accuracy should be high,

The second is to meet the high traffic flow with no sense, no contact and no need to cooperate.

Third, ease of use is strong, which can be simple and intuitive for security personnel to use.

If one element is not present, it may increase the risk of the outbreak spreading.

At the same time, AI temperature measurement scheme also began to spread to buildings, gates, access control and other low-flow entrance and exit scenarios. In addition to the synchronous promotion of sensetime, Sight-seeing, Gring deep tong and other companies, innovation Qi Zhi, Yu Pan-intelligence and other companies also launched with AI temperature measurement function of the face access control system.

While such products are valuable as part of community control, one obvious problem is the lack of facial recognition beneath masks.

Megvii’s AI temperature measurement products currently do not support face recognition under masks. “The core of the AI image technology and remote temperature measurement is to achieve a more accurate temperature measurement in the case of large crowds,” explained megvii.

Sensetime says it can achieve 85 per cent accuracy in face recognition when 50 per cent of the bridge of the nose is exposed while wearing a mask. Although some access control manufacturers have added infrared temperature measurement module, they do not support face recognition under masks. They still need to remove masks and “brush face”, or cooperate with access control card line. Visible behind the initial screening, there are still some technical bottlenecks.

Supply chain constraints
Under the epidemic, factories shut down, logistics shut down, not only the core infrared sensor shortage, even some parts are difficult to find, security enterprises encountered non-technical problems, with the assistance of local governments, just took this step of pilot.

Since the fourth day of the lunar New Year (January 28), security enterprises have promoted the pilot implementation of AI temperature measurement programs, Baidu in Beijing Qinghe station into the pilot, Kuang Vision in Beijing Haidian Government Hall and Mudanyuan subway station to pilot, Green deep Pupils in Beijing West Station into use.

After several days of trial, each AI temperature measurement system was officially put into operation around February 3. According to the data published by various companies, it can detect 15~20 people per second at high speed, and 200~300 people can pass through the temperature measurement in one minute.

According to Megvii, its AI temperature measurement system was online for six days from Feb. 4 to Feb. 10, and the actual number of temperature measured was nearly 5,000, of which more than 2,000 were measured on the first day of Beijing’s resumption of work on Feb. 10. AI has begun to support the resumption of work.

Accompanied by the rework tide, government units at all levels and public departments are in urgent need of AI temperature measuring equipment. However, at the other end of the supply chain is a serious shortage caused by the epidemic and the Shutdown of the Spring Festival. The core infrared sensors have been included in the “wartime supplies” by the government, which give priority to the epidemic areas, followed by the government, public departments, and finally enterprises.

With the help of the government, the AI temperature measuring device has been mass-produced, and about 50 sets are being supplied every day, and 80 to 100 sets will be supplied every day after February 13, said Chen Tianbo, director of products at Gling Deep Pupil. In addition to Beijing area, many other provinces and cities also put forward demands, and the company will give priority to the government and other public departments.

In order to shorten the product supply time, security enterprises dry up the “Courier”.

Chen Tianbo said, Beijing city companies directly sent car transport, operation and maintenance personnel also follow, to quickly install, the field just send aviation.

“Now the landing speed is basically in the hour, this morning (February 8) at Jinsong METRO station, 11 o ‘clock at Xiao Campi Railway station, and then to the next place without stopping.” “He said, stepping closer.

At the same time, due to the shortage of supply chain, it is difficult for enterprises to directly get infrared sensors and other core components, resulting in AI temperature measurement, face brush, access control and other products are still in pilot or supply shortage.

Innovation Qizhi said that its AI temperature measurement all-in-one access control device has been tested in some government departments, and many government and enterprise customers are interested in buying it, but it will take some time to increase the volume of products due to the tight supply chain caused by the epidemic.

A sales manager of security manufacturer Yupan Intelligence said that domestic companies give delivery cycles of temperature measuring components of more than six weeks.

Will AI temperature measurement become normal?
After this battle, there is a voice in the industry that AI temperature testing will become a basic service.

Baidu said that after the outbreak, the AI temperature measurement system can still be deployed in crowded places such as transportation stations, medical institutions and service centers to quickly detect the temperature of people entering and leaving. In the long term, crowded places will pay more attention to related capacity building.

Chen Tianbo, product director of GlING Deep Circle, said that while developing AI temperature measurement equipment, the company also extends AI temperature measurement capability to existing industry scenarios, such as face brush access control, building ladder control, and iteration of existing products.

However, due to the urgency of demand brought by epidemic control, shortage of core supply chain, and AI temperature measurement equipment mainly supplied to the government and other public departments, cost has become a neglected issue.

The market for airports, railway stations, subway stations and other high-flow AI temperature measurement equipment prices often start at ten thousand yuan, even into the integration of face access control equipment need ten thousand yuan level.

Yu Pan intelligent sales manager said, with AI temperature measurement face recognition gate hardware and software total price to 12,000 yuan. And its ordinary face brush brake only two or three thousand yuan, the price is several times higher.

Innovation qi Zhi also said that at present various components, especially infrared temperature measuring device shortage, comprehensive view, ten thousand yuan level is a reasonable price.

If AI temperature measurement becomes a basic service, such a high price will be the key to its scale.

This has to do not just with the current supply chain shortage, but with the solution itself. For example, in order to ensure the accuracy of infrared thermal imaging, Megvii and Dahua both adopt the “black body” scheme to ensure the accuracy of camera temperature measurement through the real-time temperature measurement correction of the black body, but this physical implementation method is often costly.

In contrast, Sensetime, Gerling deep pupil is using a non – black – body scheme. Chen tianbo said that it combines infrared thermal imager to measure temperature, ambient temperature, human body temperature and face temperature, forming a unique algorithm, can also control the accuracy of ±0.3℃. Software to ensure accuracy, as the market expands, or will be conducive to the formation of more competitive prices.

In addition to the price, as more and more AI temperature measurement schemes appear, or integrated into existing products, the phenomenon of the industry has appeared, but does every scene need such products? What are real needs? The industry should take a more rational view of the “new opportunities” created by the epidemic.

There are more rumors that some manufacturers will use industrial infrared sensors to measure human body temperature, industry chaos has been presented.

Accelerate the differentiation of AI security
AI infrared thermometer, epidemic prevention outbound call robot, AI diagnostic system for COVID-19, disinfection robot, epidemic prevention drone… AI is playing an important role in the pandemic.

Looking back, the epidemic is more like a training ground for technology. It tests the ability of algorithm development, solution rapid migration, supply chain integration and deployment, as well as the comprehensive ability of AI enterprises.

Although not the hardest hit areas, the epidemic has affected the business progress of AI companies to varying degrees. Under the epidemic situation, who can quickly develop products that truly solve problems and stand the test of the market, instead of blindly jamming hot spots, can occupy an advantageous position in the industry competition. In this regard, the epidemic may accelerate the differentiation among AI companies to a certain extent.

Back to the AI temperature measurement products, in addition to the participation of AI companies, there are many security companies, Hikvision, Dahua, Yuvision have appeared in the intelligent temperature measurement system, and the landing volume is wider. Its supply chain integration ability and channel ability are the direction that AI companies need to strengthen.

The growing sophistication of ARTIFICIAL intelligence is becoming a particular group of “contrarians” in the fight against COVID-19.

Artificial intelligence technologies and solutions have emerged and achieved practical results, whether it is to speed up scientific research on the epidemic in the early stage, to assist front-line epidemic diagnosis and treatment and on-site epidemic prevention and control, or to build back-end information platforms.