Home > News > Industry News > Smart security has been given .....
Browse Categories
New Arrival(26)
Hot Products(12)
Reader & Modules Series(21)
Handheld Terminal Series(14)
POS Terminal(5)
Time Attendance Series(14)
Access Control Series(6)
Sales Promotion(35)
Door Lock Series(11)
Camera Series(11)
Explosion Models
Contact us
Tel:+86-23-67305242 :
Mobile:+86-13667681778
Email:info@hfcctv.com
Web:http://www.hfteco.com/
Facebook:Huifan Technology
Contact Now

News

Smart security has been given a new mission and role

  • Author:HFSecurity
  • Release on :2020-01-03

In the past few years, great progress hasbeen made in the application of artificial intelligence in the field ofintelligent security. The iteration of computing power and the emergence ofalgorithms have indeed provided a key foundation and motivation for theapplication of artificial intelligence. However, with the deepening of AIapplications, new opportunities and challenges have come one after another,wrapped in a wave of cutting-edge technologies such as artificial intelligence,the Internet of Things, and big data. Intelligent security is given a newmission and role. 2019 , Security ushered in a qualitative change.

As an enabling technology, artificialintelligence is showing a comprehensive enabling situation in the typicalapplication scenario of security. After a few years of trial and error /groping, the industry's perception of AI has become increasingly clear andrational. What does artificial intelligence mean for security? From simple andrude face recognition in the early years, namely AI, to the computing power anddata of the three major AI algorithms recognized in the industry in the pasttwo years, the industry has generally believed that the maturity of AItechnology application is not a breakthrough in single-point technology, but Itneeds to be combined with actual application scenarios to achieve targetedsolution of bottlenecks in specific scenarios to help users improve businessefficiency / efficiency. As the AI ​​becomes more concrete, more details will be derived. At present,artificial intelligence in the security field can be said to be a combinationof algorithm + computing power + data + platform + scene, which areinterlocking and each link is of extraordinary significance.

AI concrete

From the technical level, thedevelopment of the security industry. In the past year, basically every step ofthe security artificial intelligence has achieved great progress andbreakthroughs.

Algorithm rapid evolution

First is the rapid evolution of algorithms.On the one hand, in standard scenarios, conventional algorithms such as facerecognition / license plate recognition are quite mature. It can be clearlyseen that based on mature face recognition technology, various face recognitionproducts such as face access control, face intercom, The witness verificationmachine has gradually replaced the traditional equipment, and has completelyrewritten the mode of personal identification. Similarly, the widespreadapplication of license plate recognition brings a new experience to the refinedmanagement of intelligent transportation and the unattended mode of smartparking lots.

On the other hand, with the enlargement ofthe application value of security monitoring intelligent vision sensors,relying on video, the realization of fine identification and analysis of videocontent through AI technology has become a new exploration direction for manyphysical industries. In addition to the two large-scale application areas ofpublic security and transportation, there are also many industry fields such asfinance, community, education, retail, and energy. Demand for "visual +AI" applications has also exploded. The high degree of fragmentation hasled to the emergence of various AI algorithms.

For example, a scene classification model,such as identifying a factory, a farmland, or a construction site in a video;an entity classification model, such as identifying a truck, a sports car, or ago-kart in the video; a natural camera entity detection model, such as identifyinga street side Illegal parking, unlicensed shopping, littering, etc .; humanattributes testing, such as identifying whether workers are wearing hard hats,judges wearing robes, etc .; analysis of action behaviors, such as crowdsgathering, fighting, falling down, etc. Identification analysis of behavior ...

The design of the algorithm model isdominated by scenario requirements, which involves the efficient communicationbetween the supply and demand sides for the algorithm service. For users, theyare concerned about whether the conversion of algorithm capabilities canactually improve business efficiency. For algorithm vendors, what they need topay attention to is how to efficiently obtain user needs and quickly respond tothese needs within the cost range. . In the past year, another majorbreakthrough at the algorithm level is also here. Some leading companies canprovide users with personalized AI algorithm model customization servicesthrough the AI ​​open platform, and perform model training from data annotation andimport of the labeled data. The model is deployed to the final application toform a one-stop service. The resources and capabilities based on the AI ​​open platform havegreatly reduced the threshold for AI algorithm application, including theprocess and cost of algorithm model design.

AI chip from general to special

As the core engine of artificialintelligence computing power, chips have always attracted much attention. Theindependent research and development of Chinese chips has been elevated to thenational strategic level, becoming the top priority of the 13th Five-Year Plan.

Marching into the security surveillance ofthe AIoT era, it is facing more severe pressures than ever before in theanalysis and processing of massive video data. This undoubtedly puts forwardhigher requirements for the performance of the computing power platform.Greater flexibility, lower power consumption, and cost control have become thecore demands of AI chips. Specialized chips that meet the needs of industryapplications have become a new trend.

AIoT also brings changes to the videosurveillance infrastructure. Traditional video surveillance infrastructuresface huge challenges in addressing the needs of large cross-domain videosurveillance systems. In order to meet the requirements of big data processingand timely and fast data processing, security The architecture of the videosurveillance system has gradually evolved from centralized to distributed,forming a situation of end-to-end cloud collaboration. "Cloud training,edge reasoning" has become the mainstream computing method in theindustry. Cloud chips and edge chips in the new AI chip solutions introduced bymanufacturers in the industry complement each other to continuously provideoptimized solutions for cloud edge computing power support.

At the same time, the design architectureof AI chips must rely on the scale and diversification of application scenariosand the characteristics of corresponding intelligent algorithms in order tobetter leverage the chip's computing power. Including Huawei Hisilicon andseveral other chip vendors that have entered the intelligent security industrywith the AI ​​chip business in the past few years, in the past two years, theyhave basically been doing research around dedicated scene chips. The commonpractice is to integrate the computing trends and characteristics of keyalgorithms in important application scenarios into the design of the chiparchitecture, so that the computing power and algorithms are efficientlycoordinated to greatly improve the device utilization. This type ofspecial-purpose chip can also maintain a fairly high effective utilization rateaccording to the evolution trend of the algorithm, so that customers reallybenefit from the advantages brought by algorithm innovation.

AI technology realization, data governancebecomes the key

If innovative breakthroughs in products,algorithms, and computing power are the first half of intelligent security,then when entering the AIoT era, the core theme of the second half ofintelligent security will focus on data processing and exploration of refinedapplication scenarios. The needs of users are moving from basic visualmanagement to deeper data applications. How to turn massive video data intouseful data, so as to output the "data decision" applicationcapability for various industries has become the current and future of theintelligent security industry This is the inevitable path to advance artificialintelligence from cognitive intelligence to cognitive intelligence.

To achieve this goal, of course, the coretechnology cannot be separated from technology engines such as artificialintelligence, cloud computing, and the Internet of Things. The intelligentsecurity system in this process must not only achieve basic securityprecautions, but also play a role as a visual sensor, physical Worldheterogeneous information structured processing tools and a new role inproviding data governance capabilities for industry business; another set ofdata can collect, clean, categorize and correlate all data on a large scale andautomatically, forming a unified data view and industry knowledge mapGovernance tools are also needed.

Wu Yingshi of the public security productsolution center of Minglue Technology once said that artificial intelligencemoves from cognitive intelligence to cognitive intelligence. It is necessary touse multi-dimensional perception and data management technology to connecthigh-quality data to intelligently extract knowledge and build a set. Knowledgemaps, violence mining, and perceptual data are combined into amulti-dimensional analysis and reasoning to build a decision model to transformperception into cognition, thereby establishing a data-driven big AI system.

The core value of big data lies inconvergence. At present, it is not uncommon for companies in the field ofintelligent security to focus on multi-dimensional data fusion. Judging fromthe construction of public security big data in the past two years, there havebeen several significant developments in the industry.

First, from preliminary data fusion tocomplex multi-dimensional data fusion. In addition to video surveillance data,IoT information in different dimensions, such as Wi-Fi, RFID, and electroniclicense plates, can be linked together. Through rich data types, more valuableinformation can be collided together; followed by multi-dimensional data Fusionevolved to the stage of semantic intelligence. Semantic intelligence is complexand fused intelligence, relying on intelligent speech semantic recognitiontechnology and complex data organization and fusion computing technology tocomplete the fusion analysis of multiple video and sensor data, so as toachieve a revolutionary prediction of events.

It is worth mentioning that in order toeffectively solve the "data island" problem that exists acrosssystems, platforms, and departments, and to better help the security industryform a broad and highly relevant industry knowledge map, in 2019, the originalThe data platform strategy active in the IT field and the Internet circle hasbegun to be introduced into the intelligent security industry, laying a newfoundation for industry data governance.

AIoT becomes a new "battlefield"

"The security attributes of the videosurveillance system are just the foundation. Its attributes as the visualperception system of the Internet of Things are constantly being strengthened.The combined application of visual AI and vision + IoT is creating a broadermarket space. Quantity market, which is the AIoT market. "Said ZhangXiaolin, president of Yushi Technology Strategic Marketing. "This is alsoa highly scene-oriented market. To this end, we need to break away from thetraditional chimney-type security system construction thinking and re-examinethe application requirements of the AIoT market."

Wang Gaoxin, co-founder and CEO ofHuihuifan Technology also pointed out that smart security is on the eve ofchange, and more security construction with the goal of smart city constructionis on the agenda. The city of the future will be based on the AIoT technologyarchitecture, a large sensor network + SaaS application services + cloudplatforms for various industries. The disruption lies in the connection ofdata, applications and technology. The security industry will also usher in asensor network based on AI devices such as smart security cameras, which willprovide new changes for various applications for the entire city.

Entering 2019, AIoT has already entered thecrowd. The AIoT empowerment industry itself is an ecosystem that depends oneach other and its strengths. The entire ecological chain ranges from chips toembedded software to big data platforms to user interfaces, from applicationsoftware to There are many different types of enterprise roles for hardwareequipment. Each company also has its own expertise. Due to the long industrialchain, the application field is quite rich, and the industry scene is highlyfragmented. It is difficult for one company to cover The whole industry chain.

Facing such an industrial ecology,ecological co-construction will become the core keyword of the industry.Enterprises in the industrial chain need to carry out future market strategicplanning according to their own capabilities. At the same time, they will workwith partners to build an industrial ecosystem to achieve complementaryadvantages, and mutual benefit and win-win will gradually become a significanttrend in the intelligent security industry.