2021 Security Video Surveillance O&M Market Review and Outlook
——Interview with Li Huanjia, Product Director of Hongdu Technology
Excerpt from China Security, Issue 2021.12
With the introduction of many guiding policies such as accelerating digital development and building a digital China, the major segments of the security industry have ushered in a new round of benefits. In 2021, the overall development speed of the security industry in the post-epidemic era has recovered, according to relevant data, the production scale, domestic market, corporate profitability, export market and other major prosperity indexes of the security industry in the first, second and third quarters have increased significantly, and labor demand and enterprise investment have also increased accordingly, and security enterprises have strong confidence in the future.
In the field of security video surveillance operation and maintenance services, Guangzhou Hongdu Information Technology Co., Ltd., as a technology-based enterprise focusing on the application of artificial intelligence, is a professional provider of video artificial intelligence diagnostic products in China. The company focuses on the research and development of ai, Internet of Things and other new technologies in the field of video innovation, product objectives focus on how to make AI technology and business scenarios better integration and landing; the company has the pre-research ability and landing practice ability of cutting-edge technology. The core technologies such as active video quality diagnosis technology, multi-dimensional video diagnosis technology, and ultra-high-precision human object recognition independently developed by the team have been applied in core products such as video guard, intelligent operation and maintenance platform, and visual alarm system, and the products are widely used in public security, justice, transportation, finance, education and other fields. In this issue, li Huanjia, product director of Hongdu Technology, was interviewed to analyze and prospect the characteristics and future development direction of the security video surveillance and operation and maintenance market in 2021.
"China Security": Under the influence of multiple environmental factors, what characteristics will the video surveillance market in the domestic security industry present in 2021?
Li Huanjia: After more than 40 years of development, China Security has expanded rapidly, and gradually developed into a mature industry with a huge market size and stability, and the Chinese market alone occupies nearly 60% of the global security market share. Affected by the new crown epidemic, international trade restrictions and other domestic and foreign environments, the domestic economic growth rate has slowed down in the past two years, and the development of the domestic video surveillance market has also been affected, the growth rate has slowed down, and the market focus has also undergone certain changes. First, the change of market area and users, video surveillance construction is extending from the main urban areas and developed areas to the grass-roots and underdeveloped areas such as urban villages, townships, villages, etc., and the application industry has shifted from focusing on traditional public security, justice, finance, transportation, etc. to emerging industries such as education, culture and expo, commercial housing, hospitals, and environmental protection; the second is the change in construction focus: the construction of domestic video surveillance systems is shifting from "construction - operation" to "construction - assessment - operation and maintenance", and the proportion of operation and maintenance of the monitoring system is gradually increasing.
Overall, the security industry continues to flourish, and the video surveillance market presents the following characteristics:
1. The industry will continue to maintain its development trend, and the application will expand to depth and breadth
The security industry is closely related to the economy, science and technology, the international counter-terrorism posture and the awareness of security of residents. With the continuous development of the industry, the application of traditional security fields such as public security, transportation, finance, and justice will be more in-depth; the growth rate of education, environmental protection, medical and health, safe production and other application fields will accelerate, and the application of civil fields such as intelligent buildings, communities, and residential security is also growing rapidly, and the personalized needs of users have become an important growth point. The market application extends from first- and second-tier cities to third- and fourth-tier cities and rural areas, from general demand to more detailed demand, and the market is gradually expanding
Video surveillance product applications are currently concentrated in transportation, public security law, finance, education, government, commercial housing and so on. In 2021, the construction project of residential video surveillance based on the construction of "Zhi'an Community" is increasing, which is mainly related to the expansion process of video surveillance application fields from key departments to general departments, from public to personal. The application volume of video surveillance in the general economic field is small, and the future market space is large, while the traditional industry will expand to depth and breadth.
2. The concentration of the industry continues to increase, and technological innovation has become the core factor
In recent years, the integration of technology, brand and capital in the security industry has become more in-depth, and the cooperation between enterprises has been extended from technology, products, capital, etc. to the brand and channel level, the concentration of the industry has accelerated, and the market share has further concentrated to mainstream enterprises. The competition between enterprises has gradually shifted to all-round competition in product innovation, performance optimization, product quality, corporate brand, and enterprise services.
Under the pressure of technological upgrades and price wars, small businesses will be eliminated at an accelerated pace. Judging from the winning bids of integrated projects such as safe city and intelligent transportation this year, there is no doubt that most of them are large enterprises winning the bids, and the market share of the two giants of "Haida" has exceeded half, the concentration of the video surveillance market will continue to increase year by year, the competitive pattern of oligopolies will become more obvious, and enterprises that master core technologies will be more competitive and influential in the future industrial chain competition.
The technology and concept of The Internet of Things, cloud computing, artificial intelligence and big data continue to penetrate into the security video surveillance industry, technological change and product upgrade iteration continue to expand and extend the application field of the security industry, and technological innovation has become the most critical factor for enterprise competition. The differentiated and customized needs of customers in different industries also prompt enterprises to continuously enrich their product lines and attach importance to technological innovation. The future technology development trend will be high-definition, intelligent, multi-dimensional, ultra-integrated, low cost.
3.AI application promotes the innovation and development of the industry, and cross-border cooperation deepens
With the continuous development of the security industry, the demand for commercial and civilian video surveillance equipment continues to increase, and the landing application of artificial intelligence has prompted video surveillance to derive more software and hardware products in the fields of public security, transportation, education, banking, and residential areas, and the application has shifted more to the business level. The deep integration of video surveillance surveillance systems and artificial intelligence technology has further improved its recognition rate, accuracy rate and analysis rate, and the massive data generated by the video surveillance system may derive more information data.
The continuous expansion of application fields, the rapid change of product technology, the continuous expansion of market scale, and the active entry of industrial capital are jointly promoting the continuous upgrading of competition in the video surveillance industry, and new tracks, new technologies, and new models are rapidly maturing and iterating. Intelligence makes the security attributes of video surveillance systems gradually weaken, while the Internet of Things attributes are getting stronger and stronger, and new formats such as home security, AI services in the cloud, AI algorithm suppliers, and product modules are also emerging, bringing innovation and vitality to the industry. Under the development trend of "Internet +", video surveillance equipment manufacturers actively cooperate with algorithm manufacturers, chip manufacturers, third-party software vendors, integrated solution providers, etc., so as to achieve complementary advantages in resource allocation, brand effect, and information sharing.
4. The video surveillance system becomes a management gripper, and the proportion of O&M attention is increased
China is undoubtedly the country that has done the best job in epidemic prevention management since the outbreak of the new crown epidemic. As an important management means, the video surveillance system has played an important role in the grid management at the grass-roots level, and has done a lot of work instead of manpower, such as isolation monitoring in key areas, prevention and control management, and daily mobile personnel management. Video surveillance plays an immediate role in criminal investigation, public security prevention and control, community stability maintenance, epidemic prevention and control, etc. With the improvement of video surveillance coverage, users began to pay more attention to the application of video image information, video image quality and video surveillance system application effect is closely related, is one of the key indicators that directly reflect the operation effect of the monitoring system. In this year's tendered video surveillance O&M projects, some have begun to require the use of O&M software for O&M effect assessment, or directly related to the system indicators that users pay attention to, such as equipment online rate, integrity rate, video quality, number of inspections of front-end equipment, hard disk integrity rate, image storage time, and accessability rate as specific assessment items. A district-level video surveillance operation and maintenance project in a city mentioned the retrieval of surveillance cameras at the scene of the crime, if there is no video due to equipment failure, the operation and maintenance unit will be deducted from the maintenance costs in the current period. From this aspect, it can be reflected that users have higher and higher requirements for the stability and operation quality of video surveillance systems.
"China Security": Please talk about the operation and maintenance market situation under massive video surveillance at this stage.
Li Huanjia: Video surveillance system has become an important part of urban security management, with the development of new national infrastructure, the continuous construction of smart cities, video surveillance systems towards high-definition, intelligent, large-scale, intensive development, urban security monitoring scale has grown exponentially. According to the agency's forecast, the total number of probes in China may reach 626 million by 2022. Video surveillance has gradually formed a trend of continuous expansion of scale, huge and complex types of equipment, increasingly complex network structure, many personnel involved in maintenance, accelerated pace of product upgrading, and more and more frequent system construction and upgrading activities.
Although many video surveillance systems pay great attention to the quality of front-end equipment in design, they lack management of their operation and maintenance, so failures can often not be found and solved in time, resulting in the system not being able to operate normally or even in a semi-invalid state. At present, the following problems are common in video surveillance O&M:
In 2015, the National Development and Reform Commission and nine other ministries and commissions issued a document requiring that the integrity rate of video surveillance cameras installed in key public areas should reach 98%, and the integrity rate of video surveillance cameras in public areas should reach 95%. In the same year, the Ministry of Public Security required that the operation and maintenance of safe cities in various provinces be assessed through the operation and maintenance management module of the image information ministerial networking platform, and the networking indicators, online rate, integrity rate and image quality became the standards for the evaluation of the operation and maintenance of video surveillance in the public security system. The requirements of national and local governments for video operation and maintenance are increasing year by year, and the cost of video operation and maintenance protection is also rising rapidly.
If the video surveillance system adopts the traditional manual operation and maintenance method, its passive operation and maintenance efficiency is low, the problem of missed inspection is often missed, the data statistics are incomplete, the assessment cannot be started, and it cannot adapt to the rapid growth of urban video scale and the increasing management needs of related business applications, and the operation and maintenance management will only become more and more passive. Measured from factors such as rising labor costs, technical complexity and professionalism, professional operation and maintenance led by intelligent technology tools in the future must be the development direction of the industry.
"China Security": In the future, how will the operation and maintenance of video surveillance systems and AI be integrated, and what impact will it bring to the market?
Li Huanjia: In-depth research on video diagnostic AI algorithms, as well as the integration of AI technology and big data, to achieve intelligent operation and maintenance (AIOPS).
1. The technical cornerstone of video operation and maintenance- video diagnosis AI algorithm
In recent years, visual analysis algorithms such as face recognition and object recognition have developed rapidly, but algorithms specifically applied to operation and maintenance scenarios have not received much attention. There is no breakthrough development in the industry's video diagnostic products at the algorithm level, and even the best players can only reach the highest recognition rate of 90%-95%. This accuracy rate is barely acceptable for small-scale applications, but once the number of devices exceeds the number of projects with 10,000 channels, the number of false positives is also very alarming, requiring a lot of manpower to be invested in the review process, and it is impossible to achieve automated and intelligent operation and maintenance.
At present, the face recognition and target recognition algorithms that are relatively mature and can be applied on a large scale are the main reasons for such rapid development:
(1) The model training production materials required for such algorithms are very easy to obtain;
(2) The ai basic layer supply side of the industrial support is complete, including a one-stop data governance platform, a computing hardware platform, and an AI software development platform;
(3) The supply of a complete industrial chain makes enterprises that need such technologies only need to invest less money to purchase high-quality solutions for integration.
The AI analysis algorithm of video diagnosis does not have the good development conditions of the face recognition scene. In order to produce high-quality products, there are higher thresholds in algorithm definition, data collection, model training, and accuracy evaluation.
(1) Algorithm definition: The image anomalies of video surveillance are very numerous, the type definition is blurred, and the quantification standards cannot be unified, so most image scenes require a combination of multiple algorithms, and a large number of models still need to be defined for a single image anomaly type to adapt to different scenarios;
(2) Data collection: abnormal failures are mostly accidental events, the amount of data is very small, and can not be simulated, can not mass production training data, need to have a large number of project basic data accumulation to form incubation products, the establishment of sustainable collection, continuous release of the application model in order to have the ability to continuously optimize products;
(3) Model training: abnormal fault data is mostly concentrated in common faults, abnormal image samples have very large individual differences, and the traditional model training methods and results can only meet some basic functions, and the accuracy can reach more than 90%. Without measures to deal with uncommon video quality issues and individual differences in images, system functionality and accuracy can be severely biased.
(4) Accuracy evaluation: Abnormal fault definition will inevitably be affected by subjective judgment differences, 1,000 users have 1,000 sets of measurement standards in mind, a set of standards can only approach the expectations of most scenarios, and the standard algorithm can only infinitely approach the expectations of most users, and cannot meet the needs of all user scenarios. Therefore, the video diagnosis system needs supporting technical means to deal with these contradictions and conflicts in order to effectively improve the accuracy of practical applications.
In order to improve the accuracy and adaptability of the video diagnosis system, in addition to solving the problem of continuous optimization of standardized AI algorithms, it is also necessary to pay attention to how to automate and intelligently integrate the concept of independent learning into the product use process in the process of product delivery.
The accuracy and adaptability of video diagnostic technology is the cornerstone of intelligent O&M, and Hongdu Technology provides a reliable data source for creating an intelligent video O&M system through years of research and design of active video quality diagnostic technology (Active-VQD). The new self-developed active learning framework redefines the video image quality evaluation system, intelligently analyzes, judges and alarms the quality problems existing in the real-time and video screens, allows the analysis model to be set for a single camera, realizes "thousands of machines and thousands of faces", reduces false positives and false alarms caused by different camera working environments or differences in the subjective judgment of users, and greatly improves the accuracy and adaptability of products compared with similar products, and the diagnostic accuracy rate is as high as 99%.
2. The value growth point of future video O&M products - the integration of AI technology and big data to achieve intelligent O&M (AIOPS)
The scale of the city-level video surveillance system is getting larger and larger, the networking structure of the video network is more and more complex, the possible path of a request is increased in the form of Cartesian product, and a node abnormality often causes multiple asset alarms on the network. The phenomenon that various videos cannot be called or the quality of the user is degraded is likely to be caused by multiple links. This brings great challenges to fault discovery, fault location, and fault emergency handling. In view of this situation, artificial intelligence is introduced into the automated inspection and diagnosis process, and fault location through artificial intelligence is used to improve O&M efficiency and shorten fault recovery time.
At this stage, the video O&M tool mainly discovers the failure of a single video device, but cannot accurately locate the cause of the failure and the analysis of the service impact, such as no video recording at a certain video point, which may be caused by hard disk failure, no hard disk, no video recording plan, equipment shutdown, network interruption, etc. In recent years, a variety of network monitoring, Internet of Things monitoring equipment have been launched to escort the smart city, although it can monitor the working status of front-end equipment in real time, making operation and maintenance more convenient, but far from meeting the strict requirements of accurate fault location.
Professor Pei Dan of Tsinghua University concluded that "the core challenge of intelligent operation and maintenance is: from the perspective of industry, we have data and applications, but we lack some algorithms and experience; from the perspective of academia, we have a lot of theoretical algorithms, but lack of actual data to support scientific research, and we are not familiar with the operation and maintenance scenarios." Based on this idea, we continue to study the combination point between more and more operation and maintenance scenarios and artificial intelligence by collecting a large amount of video surveillance system operation data.
There are more than a dozen intelligent operation and maintenance scenarios for video surveillance, and due to the limitation of space, this time we mainly analyze the operation and maintenance scenarios from the root cause. In large-scale video surveillance systems, completing a video-on-demand service requires multiple products, multiple nodes, multiple processes, and multiple interfaces to be called. Once a node on the video-on-demand link is abnormal, it will cause all nodes in the upstream to behave abnormally, and the monitoring system will appear as multiple products alarming at the same time. Ideally, the nodes at the point of failure and before are congested, and the nodes after the point of failure have no requests, at this time, the load and situation on the on-demand path can be visualized to quickly locate the problem. However, the actual situation is that an interface is often used by many users, a node generally provides multiple interface services, streaming media servers and other shunting devices after multiple nodes of the same type, these factors constitute a complex IT service network. Under complex networks, video
The quality is degraded, and it is difficult to cause obvious characteristics on the monitoring system. This scenario is suitable for the application of artificial intelligence, and the artificial intelligence system can well capture this feature and notify the disposal system to take action when a failure occurs and has not yet caused a large-scale business blockage. This not only reduces the emergency disposal pressure of O&M personnel, but more importantly, improves system continuity. Based on the selection of data from the O&M knowledge graph, the training model can effectively reduce the data range during model training, remove irrelevant data interference on the service network, and make the model training faster and more accurate. When training a root cause analysis model, it is mainly to find out the relationship between the location of the anomaly and the various monitoring indicators on the path.
In the future, through the rapid integration of AI and big data technologies, through global search, correlation analysis, machine learning, data modeling, etc., the accurate location of faults in video O&M should automatically analyze various fault factors for troubleshooting and finally determine the cause of the failure and automatically give maintenance guidance to assist O&M, and video O&M tools will eventually become easy-to-use tools.
From the relevant plans disclosed by the government, it can be seen that the general direction of the government's investment in public safety during the "14th Five-Year Plan" period remains unchanged, and Ping An China is still one of the main construction directions. In 2021, as the opening year of the "14th Five-Year Plan", new infrastructure, smart city construction and other related policies will promote the further application of cutting-edge technologies such as artificial intelligence, cloud computing, big data, and 5G in the security industry, and new technologies + new scenarios will become new growth points in the domestic video surveillance market, while projects such as Xueliang Project, Zhi'an Community, and Urban Old Community Transformation will also bring a certain incremental market. Similarly, the video surveillance operation and maintenance market will also rise.
As a professional third-party O&M product service provider, Hongdu Technology will work with enterprise partners to create a new ecosystem of video surveillance O&M services to better serve customers from all walks of life. In 2022, the security video surveillance operation and maintenance market will be a year where opportunities and challenges coexist, with broad prospects and a promising future!