With the continuous development of deep learning, after three waves of artificial intelligence technology, there are related industry applications in various fields such as security, finance, medical treatment, law and education.

Among them, the field of security has also been considered as one of the best artificial intelligence technology landing industry. And this is mainly due to the two major characteristics of security itself: First of all, the security industry with video technology as the core has a huge amount of data sources, which can fully meet the artificial intelligence training requirements for the algorithm model. Secondly, prior prevention in the security industry, Response, the pursuit of follow-up and artificial intelligence technical logic completely consistent.

From the current market situation, in view of the huge security market size and substantial revenue revenue prospects, it just makes it a must for many giants and start-ups. To this end, we are also curious at this node, in the face of many giants and start-ups to join the AI security market, what kind of competitive pattern, and what are the opportunities?

Security market size in the end how much?

According to the statistics of China Security Network, in 2016, the total size of China's security industry has reached 568.7 billion yuan, up 17% over the same period of last year. It is estimated that the domestic security market will maintain a growth of 15% in the next two years and the size of the security market in 2018 will reach 752.1 billion yuan . Among them, in the security industry, video surveillance accounts for nearly half of the market and is expected to maintain a CAGR of 13.4% over the next two years. The market size is expected to reach 111.4 billion yuan in 2018.

What are the typical players on the current security market?

For the convenience of everyone in the field of security can have a more intuitive understanding, we also made a brief review from the perspective of the industrial chain. At present, the participants in the upstream and downstream sectors in the industry include:

Upstream, includes video algorithm providers, chip makers, image sensors, lens modules and other core components;

The middle reaches include hardware suppliers, software service providers, system integrators and operation service providers.

Downstream, for the terminal industry applications, involving government, public security, transportation, finance, civil and other fields;

In fact, driven by technological progress, the entire security industry is also undergoing a series of phased changes. Among them, equipment suppliers are mainly in stage 1.0, solution providers in stage 2.0 and operators in stage 3.0. With the rapid development of domestic security industry, there are already a number of outstanding enterprises that have rapidly grown into the leading international security leaders such as Hikvision, UOB, Orient Power, and Suzhou Keda. From the product line of view, compared to the traditional top-down layout from the giant line, the current stage of Internet startups are more from top to bottom into the market. Typical of these are:

Traditional giant, from the channel to give full play to its scale advantage, and gradually extended to the upstream and downstream industries. On the one hand, in the key technical fields of upstream such as active layout of chips and algorithms, on the one hand, it also extends the business integration of integrators or operators downstream. It is reported that at present, Haikang and UOB two will occupy 43% market share, revenue over 10 billion, net profit of more than 1 billion.

Hikvision, as a leading security company in the country, Hikvision video surveillance intelligent has been on for 10 years, its video as the core solution and data operations Io service provider to provide global security visual management and Big data service

Dahua shares, starting from 2013 with its own industry leading edge, to meet the needs of end users, began to develop into the overall end to end to provide video surveillance solutions, systems and services provider.

Venture companies, with the advantages of technical algorithms but difficult to commercialize independently, the previous multi-way through the traditional giant market layout. Among them, from the computer vision into the typical company:

Shang-Tang Science and Technology was established in 2014 and completed the financing of B-round of 410 million U.S. dollars in July this year. Its core technologies include face technology, intelligent monitoring, image recognition and character recognition. At present, And other cooperation, but also the layout of Xinjiang market;

According to map technology, was established in 2012, completed in May this year 380 million yuan C round of financing, the core business includes smart security platform, the city data brain, smart medical and health, smart security platform through accurate face and vehicle recognition technology Has served in Suzhou City Public Security Bureau, Fujian Provincial Public Security Bureau, Guizhou Provincial Public Security Bureau and so on;

Unprecedented science and technology, was established in 2011, the end of 16 billion dollars completed C round of investment, to face recognition as the starting point, the product line includes FaceID, Face ++, intelligent real estate, intelligent security, which mainly in the field of security and follow-Knight Island , Airport cases, and railway public security cooperation;

Green deep pupil, was established in 2013, 14 years to complete the tens of millions of dollars of A round of investment products include deep pupil human eye camera, Haoyan behavior analyzer, prestige vehicle character recognition system and prestige view big data analysis platform, Security cooperation in the field of objects include the Wuhan Public Security Bureau and Tianjin Traffic Authority;

AI in the field of security have what technology applications and scenarios?

From a technical perspective, the current application of AI in the field of security mainly involves the recognition of faces and vehicles, including biometrics, big data and video structuring technologies. Among them, biometrics includes fingerprint identification, iris recognition, face recognition, gait recognition, etc. The first two are mostly used for authentication in specific scenes. As for the video structured technology, at present, it mainly combines machine vision, Artificial Intelligence, such as image processing, pattern recognition and deep learning, is also the basis for understanding video content.

Industry perspective, the current smart security in public security, transportation, buildings, finance, industry, civil and other fields have application scenarios. For example, in the field of public security, it mainly involves the application of map detection, actual combat, and pre-judgment to meet the actual needs in advance, during and after the event. In the field of transportation, in the future, through the establishment of an urban brain, AI technology can be used to analyze urban traffic flow in real time, Adjust the traffic light interval, shorten the vehicle waiting time, etc., in order to rationally allocate resources to enhance the efficiency of urban roads; intelligent buildings, the use of AI technology can comprehensively control the building's security, energy consumption, while access to buildings, people, vehicles, real-time monitoring To ensure the safety of the core area.

AI at this stage in the field of security what are the main problems?

Although the application of AI in the field of security has a very good prospect, at present the recognition of people, vehicles and objects has not reached the stage of actual application. There are still many problems that need to be constantly improved and solved. For example, the environment has poor adaptability, the data Scattered resources, limited understanding of scenarios.

1) The environment has poor adaptability. At present, due to the relative standardization of vehicles and the road environment, the recognition rate is relatively high. However, the accurate recognition of faces is easily affected by the environment such as insufficient illumination, blurred images, too small target size or mutual occlusion, So as to affect the recognition accuracy;

2) Data resources are scattered. At present, the openness and sharing of monitoring data in the field of security are relatively low. It is difficult to carry out cross-fusion analysis of multidimensional data. This makes the artificial intelligence analysis lacks effective data support and affects the accuracy as well.

3) limited understanding of the scene, the lack of effective knowledge accumulated in the field of expertise, the ability to understand video content is weak, the current intelligent analysis of mostly single-scene target detection and behavior analysis, rarely involved in a wide range of scene-related behavior Analysis, it is difficult to use for abnormal behavior analysis and risk prediction;

Future AI in the field of security what are the industry trends and opportunities for development?

1) video structured processing, the current security in the sub-industry, video surveillance market share of nearly half, which contains a large number of video surveillance data. However, at present, in the process of video structured processing, more still stays in the target recognition based on single scene of static features, and has not structured the dynamic features such as actions and behaviors and the correlation between them. In the future, if we can understand the semantics of the video, structuring the dynamic features such as the time, space, and behavior of the video will have great practical value in the later video retrieval and video analysis.

2) Intelligent front-end equipment, the previous front-end equipment can only do HD recording, intelligent analysis capabilities are relatively weak, but the current depth-based intelligent analysis technology is still more on the server for processing, once the future of video data Increased, the server bandwidth and back-end storage management bandwidth will increase the pressure, but also can not meet the security in the security intelligence, real-time, robustness and other requirements. In this way, the front-end intelligence has become the trend of the industry, that is, embedded depth learning algorithms or chips can be embedded in the front end of the device with structured information extraction, face recognition, road detection, vehicle identification and other functions. At present, such as the Intel chipset based on the IP chip algorithm, Xinbo Electronics based on the SVAC2.0 standard, and everyone's intelligence from the module.

3) Breakthrough point of technological innovation In view of the current restrictions on the accuracy of recognition of people, vehicles and objects, there are still many limitations. In the future, there are corresponding technological innovations in terms of software and hardware. Such as front-end equipment, high-definition display technology will continue to develop in the next few years, such as integrating 3D images at 4K level, obtaining more three-dimensional depth data from 2D to 3D, or real-time feedback between the optical system and the recognition system Get a clear image of long-distance objects, or use the starlight camera to bring higher contrast and better color performance to meet the high-quality night monitoring needs, etc .; in the software algorithm, the camera to capture images to give the most suitable machine To detect and recognize faces, on the one hand, a lot of algorithms need to be done to calculate the best imaging effect according to the current lighting conditions and the like, and on the other hand, a large amount of data needs to be collected in various scenarios as much as possible. Taking a parking lot as an example, when the vehicle enters the library, the light intensity of the front and the background is greatly different, and how to adjust the partial exposure parameters for the area. To put it simply, the feature extraction based on the camera's height, angle, environment, and abnormal face needs to be improved differently. Based on this, different soft scenes can be specified according to different scene needs. Hardware products.

4) Multidimensional data fusion analysis, that is, cross-fusion analysis of multi-dimensional and multi-scene data, including the recognition of the same object (people and vehicles) by different devices, and data fusion for different scenes. In the field of public safety, for example, even the error rate of 1 millionth in dynamic face recognition results in too much practical false alarm rate. To solve such problems, , Chips and other dimensions to enhance the recognition rate is not enough, the need to expand the data dimensions based on video data, such as mobile positioning data, social data, vehicle data, consumer data, through such large-scale, multi-modal data integration Further enhance the recognition effect, and reduce the error rate on the order of magnitude.

5) Subdivision Wide range of applications, benefit from the rapid development of deep learning algorithms in the field of security, intelligent security has been more and more widely used. In the era of AI + Security 3.0, in the face of the downstream demand of security video products, there will be more market space for operation and service. This will also become the future development direction of China's security industry. Take face recognition as an example, it can be widely used in public security, retail, education, finance, medical and other industries; in addition, the future can also try emerging scenes, such as the wisdom of the area to complete the items leftovers detection, passenger flow statistics and intelligent patrol Seized; wisdom business, traffic statistics and flow of people density testing.

Summary: 1. Overall, security market size is large enough, there are enough opportunities; 2. At present, the application of AI in the field of security there are still many technological breakthroughs need to be broken; 3. The future of data operations services will be larger Market opportunities; 4. For start-ups, the ability to find a breakdown of the industry scene is the key;