Seeing is real, but data doesn't lie

source：fametal.com.cn 日期：2018-09-03

If I still believe that "seeing is believing", I'm sure someone will come up with a list of counter-evidence. A physicist would tell me what a mirage is; Science and technology scholars will explain what is virtual reality (VR) and what is augmented reality (AR). Psychological experts will also explain the cognitive biases that can result from subjective assumptions. And then, of course, there's the magic, which, at the moment when the coin goes through the bottom of the cup, I don't notice that the wedding ring on the magician's hand is actually magnetic.

Data related to "people.

That's a long way off, at least in smart cities, where I still believe seeing is real. Because smart cities, especially in the field of urban governance, cannot be separated from "face watching" and the big data application behind "face watching". For example, the IT department of the public security bureau has been working on computers before. Urban governance is turning to science and technology for police force. Through the combination of big data and artificial intelligence, the monitoring of criminal behaviors in advance, in the middle and after has been realized.

Going into the theme, video cloud mode is very illustrative. Face recognition and portrait big data are playing an increasingly important role in urban governance. As far as technology is concerned, the combination of urban governance and portrait recognition involves data, computing power and artificial intelligence. Computing power is the foundation. Seven GPU servers are now deployed, equivalent to the performance of 200 previously. While artificial intelligence is the application presentation, different algorithms can support the application of different fields such as criminal investigation, traffic guidance, and violation correction.

Of course, going up to ai, going down to computing power is data. Data here refers not only to big data analysis platform, but also to data related to "people". Many fields in smart cities generate huge amounts of data every day, usually at the PB level. Not to mention all kinds of data collected by Internet of things sensors, data related to "people" alone includes population data, legal person data, Internet social data and unstructured data collected by tens of thousands of cameras. Only when the data related to "people" are integrated and coordinated among many "libraries", can the value of urban governance be generated.

As for the application of portrait big data in the field of urban governance, "people" as the data subject is the foundation of urban governance application, and the portrait recognition big data solution formed on the basis includes four parts: observation, recognition, analysis and decision-making.

Look and smell the cut portrait recognition

This principle is very close to the theory of looking, smelling, asking and cutting in traditional Chinese medicine. See? Video, picture, portrait and license plate belong to the perceptual data range of being "seen". And the purpose of "seeing" is how to recognize? Portrait recognition, the key lies in identification, including: static portrait comparison and dynamic portrait comparison. In this aspect, it involves not only the artificial intelligence algorithm, but also the front-end acquisition data, and the comparison with the back-end database.

Of course, "seeing" and "knowing" are only basic capabilities, and closely combined with smart cities, they must be demonstrated by "analysis and decision-making", which is the historical watershed of the old and new security era. In the old security era, due to the lack of "analysis and decision-making" ability, video monitoring system is only a simple image acquisition tool, which can only "see" and stop at "see".

As we enter the new security era of data intelligence, the definition of security infrastructure is no longer just "seeing", but "understanding", able to analyze and make decisions.

Specifically, the analysis modules in portrait big data solutions mainly include: individual portrait, interpersonal association and group analysis. In the aspect of individual portrait, through face recognition, we can analyze the behavior track of individuals in the past year, the appearance and disappearance place, living place, activity point and abnormal point.

For example, if you leave tongzhou at 8 a.m. every day, appear at 10 p.m. at the west second flag and go home at 10 p.m. You're probably working for a tech or Internet company. Among them, the probability of a programmer is 60%, that of a product manager is 30%, and that of a sales manager is 10%.

The individual portrait, of course, individual portraits can't rely on the camera information collected, information sources must also include online and offline correlation data, such as archive by the household registration of the ministry of public security, education, transportation, accommodations, the system data, as well as the micro letter, weibo, taobao, QQ social data such as the Internet. The complete individual portraits can only be made by combining the above data.

Through the individual portraits, the interpersonal association analysis and group analysis can be further completed. In this process, the common points of individuals and groups can be found through big data analysis: for example, traveling in the same car, accompanying, colleagues, same behavior pattern, same community, same criminal record. For example, if the drug addicts monitored by the system are present in a certain area at the same time, don't hesitate to call in the police and shut down the drug production site.

At present, based on video monitoring cloud, graph technology can provide behavior trajectory analysis, character portrait analysis, peer staff query, abnormal behavior alarm, abnormal site alarm, population density analysis and other applications. Therefore, the smart city is entering the era of "face watching", and the foundation of "face watching" is the computing power and the data analysis ability based on "people".