The data shows that the scale of my country’s security market has increased from 324 billion yuan in 2012 to 660 billion yuan in 2018, with an average annual compound growth rate of 12.6%. Among them, the scale of the smart security market is close to 30 billion yuan, and it is expected to reach more than 100 billion yuan by 2020. Further refinement, the market size of the AI security industry, one of the branches of intelligent security, reached 13.5 billion yuan in 2018, an increase of nearly 250% compared to 2017.
Another data shows that AI will promote the market size of the security industry to approach one trillion yuan in 2022. It is conceivable that AI has strategic significance for the security industry. Only gradually, AI security seems to have fallen into a vicious circle. Whenever AI security is mentioned, people often think of various surveillance cameras first. When it comes to products, does AI security only have surveillance cameras?
In the field of AI security, surveillance cameras occupy the largest hardware market
According to functions and uses, security products can be divided into monitoring, detection, protection, etc., and video surveillance occupies a large part of them. According to previous data, in my country’s security equipment market, the market share of video surveillance products has reached more than 50%, and it has become the core product for building security equipment systems. Regardless of the traditional security market, or today’s smart security and AI security markets, the status of video surveillance cameras has not changed.
Market research agencies released survey data in 2017, saying that in 2016, the number of surveillance cameras installed in China’s public and private areas (including airports, railway stations and streets) has reached 176 million, and it is expected that this number will increase in the next three years. The annual growth rate doubled to 626 million. In addition, IDC has also made a forecast on the deployment of surveillance cameras in China, saying that by 2022, the deployment will reach 2.76 billion. The application scenarios cover public security, traffic management, emergency command, disaster prevention and early warning, and emergency repair of municipal facilities. etc.
At the same time, at the company level, the product direction of its security business is mostly around image data, including surveillance cameras, monitoring platforms based on data collected by cameras, big data platforms, and a series of chips developed by enterprises. . It can be seen that in the field of AI security, what most companies want to do is to empower surveillance cameras and convert passive security into active security at the monitoring level.
As far as the field of surveillance cameras is concerned, the hardware manufacturing level is almost divided by Hikvision, Dahua, and Uniview, and the algorithm and software aspects are occupied by giants such as Huawei and new start-ups such as CV Four Tigers and Bitmain. Already quite “crowded”.
It is undeniable that surveillance cameras have become the “darling” of AI security. A question needs to be raised here, what else does AI security have besides cameras?
From software to hardware, AI security is not just a camera
At the Amusement Expo site, in addition to the cameras hanging on the wall that make patients with intensive phobia uncomfortable, there are also some AI security “treasure products” hidden in it, and they seek development under the strong offensive of surveillance cameras.
· At the hardware level, technology has more carriers
In the past, people often said that “seeing is believing, hearing is believing”. Although sometimes what people see is not necessarily the truth, the act of “seeing” is regarded as an important means in the security industry, especially in the prevention and monitoring link.
Taking criminal investigations as an example, in cases where witnesses participate in the investigation, the police sometimes restore the suspect’s facial portrait as much as possible through the witness’s dictation. This portrait will be used as a biometric feature to help the police be more accurate, Apprehend suspects faster. Compared with voices, fingerprints and other data that are difficult to identify and may have defects and flaws, in terms of finding people, most of the time ultimately rely on face feature data, and face recognition also has an advantage in speed. In the field of AI security, visual artificial intelligence technology is quite necessary, but the final carrier is not only the surveillance camera, which has already become a “red sea” product field, it also has more combination possibilities.
Taking intelligent security robots as an example, this is another AI security hardware that many manufacturers pay attention to besides surveillance cameras. Compared with surveillance cameras, security robots are more of an AI technology complex. In addition to the basic visual artificial intelligence technology, its “body” can also be equipped with sensors such as smoke detectors and odor detectors, as well as voice artificial intelligence products. Offensive weapons such as strong sound dispersion system.
In terms of function, although the surveillance camera initiates an active warning after identifying the target person, what it essentially provides is a single visual recognition and monitoring. In contrast, in addition to providing visual recognition and monitoring services during autonomous patrols, security robots can also provide functions such as sound collection and recognition, instant voice, dangerous gas recognition, and dispersal. In addition, in terms of landing scenarios, airports, warehouses, parks, hazardous chemical enterprises, banks, commercial centers, communities, etc. can all become inspection sites for security robots.
Previously, Huatai Securities had estimated that the total demand for the domestic inspection robot market from 2018 to 2020 was about 47.7 billion yuan, with an average annual demand of about 15.9 billion, respectively corresponding to 9,000 substation inspection robots, with a market space of 7.2 billion, and 8 distribution station inspection robots. ．10,000 units, the market space is 40.5 billion yuan. Among them, power stations and distribution stations are just two of the many scenarios for security inspection robots. It is conceivable that there is a large market behind this industry. At present, there are more than 30 domestic companies engaged in this field, but the industry is far from exploding, which means that there are still more opportunities in the market.
In addition, in terms of portability, surveillance cameras are fixed, and security robots are temporarily difficult to carry around anytime and anywhere due to their large size and regulatory issues. Portable AI security equipment has also become a necessary device.
In this regard, the currently possible product is AR glasses. After wearing it, once a certain device recognizes a suspicious person, the AR glasses will automatically display the gender, identity and other information of the suspicious person in front of the police. Relevant security personnel can also see cables, pipes, pipes, etc. behind the wall or under the floor through the AR glasses. It will also become possible for the police to simulate and reproduce the operation of other equipment, and for firefighters to monitor and judge in real time in the fire scene.
Like security robots, due to factors such as network transmission delay, lens imaging technology, and optical module immaturity, AR glasses are far from reaching large-scale maturity, and currently only have small-scale applications. At present, there are still a few companies that launch AR glasses in the security field, and the market is also developing slowly due to the decline in popularity. For the AI security industry, AR glasses are also a “big blue ocean”.
Under the strong cover of surveillance cameras, there are still many industries such as security robots and AR glasses that have huge markets and have not been paid attention to.
· At the software level, data barriers still need to be broken
It should be noted that the realization of all the above product functions requires the cooperation of algorithms, computing power, data and intelligent networking.
Right now, after several years of development, optimization and running-in, all kinds of intelligent algorithms have become relatively mature, and it is no problem to meet the basic needs, and the problem of computing power has also been solved because of AI chips. However, the remaining data has become a problem.
Take the public security, an important organization in the security field, as an example. As early as the beginning of 2018, the Ministry of Public Security formally established the National Public Security Big Data Work Leading Group, announcing that it will vigorously implement the public security big data strategy.
As a vertical industry, data in the public security field also has a “common problem”, that is, AI applications can only be carried out based on structured data, while those unstructured and semi-structured data have not been able to play a real role; There is also a lack of correlation between public security, between various police departments, and between various data, which also greatly reduces the capabilities and effectiveness of prediction, early warning, and prevention.
How can these problems be solved? Need to rely on industry knowledge graph.
The establishment of a knowledge map is a very difficult task, which involves steps such as data collection, data cleaning, data definition, relationship extraction, knowledge storage, and associated calculations. It is different from general knowledge maps and builds professional knowledge maps in vertical fields. more difficult.
And once the knowledge map is built, the utility it brings is also visible to the naked eye, realizing the in-depth integration and transformation of actual combat experience and technical algorithms, making police work more intelligent. To give an example, in order to lock down a criminal suspect, the police often need to call multiple systems to query data, and then make analysis and integration. This process is cumbersome, time-consuming and labor-intensive. The time and energy consumed will be greatly reduced, because after the request is made, the system can retrieve relevant case information by itself, or generate a report, or present the correlation between various clues in a visual way, etc.
According to data from iResearch, in 2018, the application rate of tool-based public security knowledge maps was 30%, and the rate of platform-based knowledge graphs reached 10%. At the moment, the first step of building a knowledge map, that is, “data collection” is still not well completed. As a result, the subsequent steps will also encounter some obstacles and difficulties.
For AI security, computing power, algorithms, data and connections are indispensable.
According to the “2018 Global Security Top 50” list released by a&s, there are 32 companies specializing in video surveillance, and Hikvision, Dahua, etc. are also involved in this field. It can be said that video surveillance has become the mainstream of the security industry. It is no wonder that in the process of intelligent upgrading, surveillance cameras will become the market entry chosen by many AI companies and the key direction of security business.
But from the perspective of market and demand, surveillance cameras are not omnipotent, and there are always places that it cannot take care of, and these are the opportunities left to manufacturers. If the development is smooth enough, perhaps these industries will also be able to “shoulder-to-shoulder” surveillance cameras in the future.