[ Security Exhibition Network Perspective Tracking ] The development of the security industry in the past 20 years has gone from analog to digital, from standard definition to high definition, and has now entered the era of artificial intelligence. The goals of digital and high-definition change are clear. There is no hesitation and repetition in the industry, and almost the starting gun is fired. The effect of commercial landing is also significant. The security industry reshuffled, changing the small and scattered state, and forming a pattern of security industry head enterprises with strong landing and innovation capabilities. With the aforementioned good industry foundation, and relying on the natural advantages of security with video as its core, security has become a good choice for the implementation of artificial intelligence.
According to the statistics of iResearch, in 2018, China ’s artificial intelligence empowered the real economy's various industries, and “AI + security” accounted for more than 50%. The growth rate of artificial intelligence in the security industry has also continued to accelerate, and the proportion of AI in security companies has continued to increase. According to iResearch, in 2018, China's AI + security software and hardware market reached 13.5 billion yuan, some The security business ’s AI business accounted for about 8% of total revenue, and some typical AI companies ’security business accounted for nearly half of their operating income. In 2018, the penetration rate of AI in urban public security reached 2.6%. It is expected that the market will still maintain a high growth rate in 2019. By the end of the 13th Five-Year Plan period, the growth rate will start to stabilize in 2020, and the market size will reach 45.3 billion yuan (11% of urban public security AI penetration rate). It is expected to exceed 70 billion yuan (the penetration rate of urban public security AI reaches 25%).
Although artificial intelligence has made a good start in the security industry, it has a good landing and a fast growth rate, but it does not mean that the development process of artificial intelligence in the security industry is smooth and the road ahead is smooth. There are many challenges and problems to be solved. The following describes the current problems and challenges from the perspective of engineering challenges, lack of standardization, and fragmentation of AI requirements.
Looking at the current security + AI landing scene, the problem of engineering properties is a necessary condition. The front-end engineering properties include layout, workability, poles, installation, fill light, debugging, and so on. The back-end engineering properties are mainly Refers to data warfare. For example, the application of ITS is a typical AI landing scene. The earliest artificial intelligence in the security industry started from ITS. The front-end bayonet and electric alarm and other equipment were clearly defined with engineering attribute parameters. 5. What is the distance from the snap point to the pole, what kind of lens is equipped, how many lanes are captured, how much is the intensity of the fill light, what is the installation irradiation angle, what are the debugging specifications, etc., there are strict workmanship forms and implementation specifications. A system installed and debugged according to such engineering attributes, coupled with back-end data technology and warfare methods, can meet customer needs as much as possible. The entrance / exit parking lot vehicle management system derived from ITS, the clear engineering attribute definition, and the back-end data operation and maintenance platform are typical cases derived from to G end to B end.
Another example is the face recognition camera, which requires installation height, face pixel size, fill light, and so on. Conversely thinking, if the possibility of implementing these engineering attributes is left or weakened, can the landing of artificial intelligence in the security industry continue to exert its advantages? Yushi believes: First, it emphasizes that engineering is essentially correct. Engineering means the accurate definition of scenes and the implementation and duplication of artificial intelligence. Second, the continuous improvement of algorithms and the continuous enrichment of scene materials for algorithm training. The adaptability of generalized artificial intelligence gradually weakens its reliance on engineering attributes; again, customers adjust their needs and expectations for artificial intelligence, such as using two pan-intelligent cameras instead of a professional intelligent camera with strong engineering attributes. The capture rate may achieve the same effect, but it reduces the difficulty of project implementation and the overall TCO.
Lack of standardization
The development of an industry is inseparable from the formulation of standardization. Standardization is a necessary condition for modern large-scale production. It can improve efficiency, scientific management, improve information flow, and incubate innovation. Standardization has a fundamental, supportive, and leading role in the development of artificial intelligence and its industry. It is both a key player in promoting the development of industrial innovation and the commanding height of industrial competition. Developed countries in the world have been fighting for dominance in the new round of international competition. They have adopted plans and policies around artificial intelligence, deployed artificial intelligence core technologies, top talents, standards and specifications, and accelerated the development of artificial intelligence technology and industry. Major technology companies continue to increase investment in capital and manpower to seize the commanding heights of artificial intelligence development.
The security industry currently lacks the industry technical standards for artificial intelligence to unify and regulate the implementation and continuous innovation of artificial intelligence in the security industry. For example, the computing power of artificial intelligence chips is nominal, currently there is no unified standard to regulate it. Each manufacturer has its own nominal test evaluation system, and horizontal comparison standards between chips are missing. If you want to evaluate the true computing power of a chip, the current method can only be compared with the actual measurement of the algorithm by manpower. Another example is the face recognition system and image search system. Each manufacturer has its own algorithm and model. If a customer uses different algorithms, there will be problems in the research and judgment on the back-end data side. Data generated by different models of different manufacturers They cannot communicate with each other, and can only be analyzed in a fragmented manner, which will cause great trouble to data analysis and equipment expansion. In January 2018, the establishment of the National Artificial Intelligence Standardization General Group and the Expert Advisory Group was held. At the meeting, the National Standardization Management Committee announced the establishment of the National Artificial Intelligence Standardization General Group and Expert Advisory Group, which is responsible for overall planning and coordination of the artificial intelligence standardization work in China.
Fragmentation of AI requirements
The great popularity of artificial intelligence has gradually increased people's expectations for artificial intelligence to solve pain points in various industries. Before the video surveillance era ushered in the era of artificial intelligence, it was mainly checked and viewed manually. Because of the popularity of artificial intelligence, people will put forward a variety of intelligent needs: chemical plants propose to detect raw material leaks, pet stores propose to detect pets, kitchens propose to detect irregular operating behaviors, and so on. The strong AI demand contrasts with the difficulty of fragmented AI landing. To solve the difficult problem of fragmented AI requirements, computing power, algorithms, and data are three important factors. At the same time, some people are already trying to use open training platforms to solve the problem of fragmented demand satisfaction. The training data and application scenario issues that are closely related to the needs are given to the demand proposers, and the platform takes advantage of computing power and algorithms.
Overall, security provides a good soil for the landing of artificial intelligence, and artificial intelligence provides a broader stage for the development of the security industry. With the empowerment of artificial intelligence and the connection with video as the core technology, the boundaries of the security industry have become blurred, and cross-borders from smart security links to smart finance, smart retail, smart education, smart manufacturing, etc. More and more frequently.