Four "persistent problems" in traditional communities need to be solved urgently

As we all know, there are many participants in smart community construction, including technology providers, equipment manufacturers, integrators, real estate developers and property companies. Based on the different attributes and positions of enterprises, each enterprise provides smart community solutions from different dimensions and levels to achieve the intelligent upgrade of traditional communities.

Xu Biao, general manager of Shanghai Trendsee Information Technology Co., Ltd., a technology manufacturer that has focused on the research and development of AI algorithms for 20 years, shared with us the application value of video-based AI algorithms in the construction of smart communities and how Trendsee is deeply involved in smart communities. In order to provide inspiration and thinking to the industry.

Four "persistent diseases" in traditional communities need to be solved urgently

The community is the epitome and component cell of the city. Currently, the urban population is dense and the floating population is increasing, which has triggered urban management issues such as social security and prevention of key areas. Urban management issues are also mapped to the field of community management. It can be said that smart communities are a product of China's urbanization trend and an extension of smart government affairs, public services, security monitoring and other intelligent systems involved in smart cities.

At present, there are four major management problems in traditional communities: 1. Personnel management and control problems. On the one hand, there is a large number of outsiders in the community, and the phenomenon of "group renting" in the community has been repeatedly banned. It is difficult for property and other departments to come to the door to check and arrange. After the ban, it will "resurgence". On the other hand, On the one hand, the monitoring of key personnel in the community (elderly people, disabled people, and "five categories" personnel) requires manual surveillance and inspection, which is timely and low; 2. Public area management problems, such as random parking of vehicles or piles of garbage in important areas such as fire escapes , In addition, mobile unlicensed vendors and shops occupy the road for operation, resulting in higher labor costs for law enforcement; 3. Perimeter anti-theft and security facilities are not in place, making it difficult to protect the lives and property of residents in the community; 4. High-altitude objects are thrown from time to time. There are few witnesses, the throwing time is short, and it is difficult to obtain evidence and be punished.

The construction of smart communities requires strong support from government policies, and it also requires the government to lead and support enterprises to jointly formulate industrial development strategies and thinking. It is understood that the Shanghai government officially released the "Three-Year Plan for Refined Urban Management" in June 2018, which clearly stipulates that the construction of smart communities will be completed within three years (2018-2022), and this is a government-funded and led focus on The smart community business model of smart security system construction will greatly promote the pace of Shanghai smart community construction and has important reference significance for the construction of smart communities across the country.

Revitalizing community video data is the core value of AI

The basic hardware environment of a smart community includes various sensors and smart devices installed in the home (smart home category), communication network hardware facilities, and various cameras and locators for video surveillance. Among them, video surveillance inside and outside the community belongs to the pan-security category, which is related to the safety of people and assets in the entire community and is the most basic part of the entire smart community construction.

However, at present, video surveillance in most traditional communities is generally only used for video display and is mainly used for real-time monitoring and post-processing. The management efficiency is relatively low, resulting in the management and control problems faced by traditional communities mentioned above, and it is impossible to control the entire community. to safety and security.

Xu Biao said that smart community is an industry covering comprehensive technology applications such as artificial intelligence, big data, Internet of Things, and network communications. Its core value is mainly reflected in the innovative applications of artificial intelligence and Internet of Things technology. The use of Internet of Things technology can transform traditional communities into Various types of equipment within the system are intelligently transformed or new basic sensing equipment is added. The status of each equipment is controlled through the back-end cloud platform and various front-end data are collected. Finally, AI algorithm technology is used to intelligently analyze the collected data, thus greatly improving the traditional The application value of video surveillance can more efficiently ensure the safety of the community and promote refined management of the government and property management.

Among them, face and specific behavior recognition analysis has important application value for community early warning and comprehensive management services. Relying on the data cloud platform, a powerful community face database and comprehensive community management can be built with video data such as residents, real estate, access control, and early warning. Database, by combining face capture, access control video and face recognition, suspicious persons are found and the information is immediately pushed to the public security department, which can effectively improve community security; using the security early warning function of access control, when fire problems occur, timely Push information to fire departments, etc.

Specifically at the human algorithm level, multiple human detection methods are integrated under the guidance of camera calibration parameters, such as face detection, head and shoulder detection, target shape and size detection, motion shape detection, and detection based on static images and deep learning. When a human target appears in the camera, the system can identify the human target and assign a unique ID to continuously track the moving target. While ensuring no false alarms, false alarms are effectively reduced and work efficiency is greatly improved. Xu Biao said that this kind of video analysis of faces and specific behaviors can effectively solve traditional problems such as personnel control, perimeter theft prevention, public area management and high-altitude throwing objects in the community.

For example, face capture and capture are carried out at corridor access control and entrances and exits to solve the problem of regular monitoring of special groups such as group renters and lonely elderly people; in areas where vehicles are prohibited from staying, when the vehicle stays for more than 300 seconds, or in areas where cross-district business operations are prohibited , temporarily setting up tents or tables on the road for more than 600 seconds, or temporarily setting up stalls in crosswalks for more than 600 seconds, or placing garbage or garbage bags in the area around trash cans for more than 5 seconds. When the above-mentioned conditions are detected, a signal reminder will be generated, and other irrelevant factors will not be identified, effectively solving various problems in public area management.

In terms of perimeter security prevention and control, areas of any shape can be set up for defense, and the alarm can be triggered no matter when the intruder enters or leaves from any direction. It also supports setting parameters such as the type, size, and minimum intrusion time of the intruder, and the alarm screen Real-time viewing, automatic capture, automatic alarm pop-up, and system linkage between the security center and the owner's mobile phone. In addition, by applying artificial intelligence technology, the occurrence of high-altitude parabolic phenomena can be automatically detected and alarmed in a timely manner.

Deeply explore the integration algorithms and business needs of subdivided industries

In the past three years, artificial intelligence technology has made major breakthroughs, but it is still in the era of weak artificial intelligence. It is difficult for any kind of AI technology to be applied across industries. Xu Biao emphasized in the interview that whether it is front-end AI chips or back-end AI Algorithms currently have strong industry attributes. To realize the industrial application of AI technology, AI companies need to have a more pragmatic attitude and should focus on the application of vertical industry segmentation scenarios while reducing the two major indicators of omissions and false positives, thereby Promote the scale of AI technology industry applications.

When talking about the current competitive situation in the commercialization of AI technology, Xu Biao said that there are many domestic AI technology innovation companies, but the technology differences are relatively large. Moreover, everyone uses different thinking to solve technology commercialization problems in different tracks. At present, Mainly using two tracks, special technology and general technology, showing a situation where a hundred schools of thought are contending. AI technology companies are far from competing on the same dimension.

In addition, no matter which track they choose, AI companies must first solve the actual needs of users and bring certain application value and innovative services to the industry. This is the inevitable path for the development of AI companies.

Moreover, he pointed out that the entire artificial intelligence industry is still in the pilot stage, and industry application requirements have not yet been determined. It is still in the stage of business integration between platforms, and every technical depth should be as practical as possible. need. However, when future application requirements are standardized, the integration of AI technology may become an inevitable trend.

The troika of rapid development of artificial intelligence is algorithm, data and hardware computing power. The algorithm is in the middle layer, which means it needs to rely on massive industry data to train the model, and it must also rely on hardware computing power to increase processing speed. As AI chip manufacturers gradually expand their chip business into the algorithm field, some analysts in the industry believe that AI algorithm manufacturers will be in an embarrassing situation and their survival space will become smaller and smaller.

In response to this view, Xu Biao believes that algorithm chipization is an inevitable trend in the future, and industry applications also need to be promoted by chip scale. However, the most critical issue is whether the function and accuracy of the algorithm module embedded in the front-end chip can meet actual business needs, otherwise it will lose any meaning.

At present, the camera's ARM cannot carry complex algorithms. At the same time, cost factors also need to be considered. Different business scenarios have different algorithm requirements. Currently, from the perspective of smart community applications, back-end algorithms have very obvious advantages. In addition, when computing power is insufficient, AI algorithm manufacturers have two solutions. One is to cooperate with more powerful chips, and the other is to optimize the algorithm so that the algorithm can achieve the same effect when running on low-end chips. In other words, Find the best balance between cost and performance.

In addition, the choice of industry segments is of great significance to AI companies. Xu Biao said that judicial prisons, retail stores and smart communities are important industries that are trending. In the future, they will delve into these subdivided areas with a more pragmatic attitude and continuously optimize independent algorithms to perfectly connect AI intelligent video analysis with actual business scenarios. In the field of smart communities, the management of public areas in the community and the area along the street at the community entrance is a top priority. Trendsee uses face and behavior video analysis solutions to solve the regional management and control problems faced by traditional communities. In terms of market strategy, on the one hand, we cooperate with integrators to provide complete product solutions and data cloud platforms; on the other hand, we cooperate with platform providers to embed algorithms into subsystems or modules into the platform.