AI in Security
In the rapid development of science and technology today, people are enjoying the sense of security brought by security monitoring. In addition to real-time recording of ongoing events, security monitoring can also analyze the monitoring video and extract information (such as license plate, face and movement analysis), so as to push video warning and realize prevention. Can play a "nip in the bud" role.
In recent years, with the rapid development of artificial intelligence and big data technology, artificial intelligence and big data technology have gained unprecedented popularity in the industry market. They have taken the lead in solving industrial pain points in areas with high data availability.
Security and defense field is the natural scene of perfect landing of artificial intelligence
1. Big data is an important support for ai analysis and prediction and self-improvement, and the security industry has a large amount of video data with rich levels, which can fully meet the requirements of ai for algorithm model training.
2. In the field of security monitoring and control, there is a logical need for pre-prevention, in-process response and post-tracking. Artificial intelligence can provide new solutions to this problem.
3. According to statistics, the security market has maintained double-digit growth from 2011 to 2016, and is expected to continue to grow from 2016 to 2020. The annual compound growth rate is expected to be 7.6%. In view of such a large market size, artificial intelligence has long been "ready to move"!
In terms of technical demand and market size, security and defense field is the fastest and best "landing site" of artificial intelligence. Although traditional security monitoring realizes the recording function, it cannot accurately identify people, objects and scenes in video. And intelligent video monitoring system with vision can completely change this situation.
Intelligent video monitoring system category
1. Identification of people and objects: it is mainly to identify the content concerned by the monitoring system, including face recognition, license plate number recognition, vehicle type recognition, ship recognition, traffic light recognition and so on. Identification technology is often used in road monitoring, financial banking, channel management and other industries, mainly to provide customers with identification records and hierarchical management basis.
2. Recognition of the movement track of people and objects: at present, there are many subdivisions, including virtual warning line, virtual warning area, automatic PTZ tracking, number statistics, traffic flow statistics, the appearance and disappearance of objects, sudden running of people, sudden gathering of people and so on. Such technology, in addition to the number of statistics, is generally to judge a process, once found abnormal circumstances, such as someone entered the warning area, the northeast corner of the square someone quickly gathered, and so on, an alarm message will be issued to remind the on-duty monitoring personnel to pay attention to the corresponding hot spots. At present, such functions are mainly used in ping 'an city construction, commercial monitoring and other industries.
3. Judgment and compensation of environmental impact: normal monitoring function can be realized in complex background environment. The environmental impact mainly includes bad weather such as rain, snow, fog, low illumination at night, camera shielding or deviation, camera jitter and so on. The application of intelligent video monitoring system technology can realize the normal monitoring function under the bad video environment. When video is not clear due to environmental impact, it should detect the person in the picture as soon as possible, or issue an alarm after judging the camera deviation. The key technical point of this kind of function is that it can output the information of intelligent analysis stably in various application situations, and minimize the impact of the environment on video monitoring.
With the development of high definition, deep learning, cloud storage, GPU and Internet of things, there are mainly two ways for ai to enter the security field:
1. From product to technology: enterprises in the field of traditional security constantly accelerate the intelligent process of hardware and software products;
2. From technology to products: artificial intelligence enterprises, which are good at algorithms, actively layout the fields of terminal application industry, such as finance, public security, building park, etc.
The application of artificial intelligence in security field
1. Public security field
Intelligent monitoring can analyze massive data in real time and extract effective clues, lock the track of criminal suspects or vehicles, and complete a series of tasks such as tracking, arrest and rescue. This process involves three layers of application, namely, graph detection, actual combat and prediction, so as to meet the practical needs of pre-prevention, in-process response and post-investigation, and effectively prevent and defuse all kinds of security risks.
Intelligent monitoring and analysis of density distribution and change trend of people and vehicles for dynamic monitoring, by adjusting the traffic light interval reasonable allocation of resources, improve traffic efficiency, provide favorable guarantee for the smooth travel of the public.
3. Intelligent buildings and park areas
The use of biometric identification technology in the access control and attendance system can not only distinguish between staff and non-staff, but also play a role in reducing building energy consumption.
4. Retail sector
Provide managers with remote visualization, customer data collection and analysis, POS cashier supervision of the shop tour.
5. Application in civil security
Intelligent monitoring can provide differentiated services for each user. Take family security as an example, when there is no one at home, the intelligent monitoring system enters the distribution mode; When there is an exception, issue an alarm and inform the owner in time. Although the prospect of intelligent security is broad, but the domestic foundation is still relatively weak, in the application process there are still many obstacles and difficulties.
Artificial intelligence in the field of security monitoring and control of the four pain points
1. Poor adaptability to the environment
The identification of video content by artificial intelligence is susceptible to environmental changes such as lighting conditions, weather factors, image quality, target size and ground object shielding.
2. Data island and dispersion
In the traditional security system, the data of each platform system is less open and Shared with each other, so it is difficult to carry out multi-dimensional data fusion analysis. Taking face recognition as an example, in order to improve the accuracy of face recognition, it is not enough to simply improve the computational power of the algorithm, but also to expand the latitude of the analysis data, such as the data that can be collected, such as location, social contact, vehicle, consumption, etc., through such large-scale multi-modal data integration, the purpose of target tracking and analysis can be achieved.
3. Limited scenario understanding
Reason 1: insufficient accumulation of professional knowledge and experience;
Reason 2: in the process of video structuring, intelligent monitoring still stays in the single scene environment based on static features, rarely involves the analysis of correlation behavior of a wide range of scenes, and does not make structural processing of dynamic features such as actions and behaviors and the correlation between them.
4. Lack of self-improvement ability
At present, artificial intelligence does not have the ability to grow up by itself, and it can only make independent analysis according to the set conditions, rather than improve itself according to the ability to share and accumulate experience.
Based on the present, looking into the future, although the road of intelligent security is very tortuous, but with the strong support of the government, open sharing of data, algorithm computing power continues to improve, artificial intelligence and security will be fully integrated, intelligent security era is accelerating.