In reality, more intelligent needs are being tapped

Thanks to the continuous maturity of the Internet of Things, big data, cloud computing and artificial intelligence technologies, the wave of intelligent upgrade based on visual perception has arrived, more and more scenes around are being occupied by AI, and more intelligent and convenient services are being developed. be excavated. Taking AI-enabled public security as an example, video security equipment based on portrait recognition and image recognition has become the focus of current public security work. By making full use of portrait recognition and structuring technologies, it is possible to centrally collect and manage all personnel, vehicles and non-motor vehicles in the public environment. Another example is in airports, customs, stations and other places, using face recognition for control, or providing convenient channels through face verification; in community scenarios, face recognition, image recognition and other technologies can provide face recognition, perimeter prevention, and stranger alarm. and other applications; while in the campus scenario, there are applications such as portrait roll call and face-scanning verification in the examination room.

While subverting the traditional values of the security industry, AI is also promoting the transformation, upgrading and innovative development of various industries, further enriching the imagination of intelligent business needs, and catalyzing the emergence of a large number of intelligent needs in the form of fragmentation and scene.

Traditional smart cameras face many dilemmas

However, with the emergence of a large number of intelligent application requirements, as the core equipment for visual perception, the smart camera, the terminal for collecting massive data of intelligent security, faces many problems and challenges. First of all, the function is solidified. Most of the current smart cameras are configured based on specific scenarios. In a single scenario, only one algorithm is often equipped. For example, in urban road bayonet, only the license plate recognition algorithm is often configured, and this algorithm cannot be changed in the future. Yes, the "sparrow pole" phenomenon common at many intersections is caused by this. The second is that the intelligent algorithm in the camera that has been configured cannot evolve. With the accelerated integration of big data and artificial intelligence, the current iteration speed of smart algorithm updates happens at all times, and the accuracy may be improved every month or even every week, while the algorithms built in traditional smart cameras are difficult to upgrade and iterate, even if The upgrade process is also very cumbersome and faces the risk of data loss. At the same time, related business applications must be suspended. Once a key video is missed during the suspension process, it will bring immeasurable losses and troubles to the business.

In addition, it is worth paying special attention to the fact that the current intelligent promotion of the security industry is still in the first half. According to the prediction of relevant institutions, the current penetration rate of domestic AI cameras is less than 5%. The non-smart cameras that have landed in the construction of "Safe City", how these devices can effectively recycle the old is worth the industry's deep thinking. At the same time, with the improvement of people's awareness of environmental protection, traditional cameras and some current smart cameras have obvious technical defects. The serious light pollution caused by white light supplementation of cameras at night has become a common concern in the industry.

Technological progress and computing power increase bring more possibilities to the industry

In the past, due to the limitations of space, power consumption, cost and other factors, hardware computing resources have been unable to meet the perception intelligence of security front-end acquisition equipment, and can only run relatively simple algorithms with low real-time requirements. With the rapid growth of massive security data, in order to further improve the data processing effect and meet the high real-time requirements of special industries, more and more intelligent computing has shifted from the back-end platform to the front-end collection equipment. Especially with the development of chips, the reduction of the volume of chips specially designed for vision, the reduction of energy consumption and the enhancement of processing power, it is possible to place AI chips in front-end cameras for good applications.

By advancing computing power and algorithms, the front-end equipment continues to have powerful video image acquisition capabilities, as well as preliminary data analysis and storage capabilities. Migrating to the edge also shares the pressure that the transmission system has been unable to keep up with the development, as well as the computing and storage pressure of the cloud center, further improving the speed of video analysis and processing, and meeting the actual business needs of the public security and other industries. With the application of high-precision algorithms and the continuous popularization of computing power, today's cameras are not only visual collectors, but also have the ability to pre-process and analyze data, and have gradually become the "eyes" that perceive everything in the Internet of Things world. Intelligent services in all industries provide rich perception data.

Cameras embrace change, software-defined has unique advantages

Facing the problems that need to be solved urgently in the development of the industry, most enterprises have accumulated profound accumulation and technical advantages in the fields of chips, artificial intelligence, cloud computing, big data, active safety, Internet of Things, etc., focusing on the field of "AI vision" technology, and launched a Vitality Camera - Software Defined Camera. It is understood that the software-defined camera adopts an open design framework. On the basis of platformization and standardization of hardware resources, the camera's software and hardware are decoupled, algorithms and software are decoupled, and the lightweight container technology is used to build multi-algorithm-oriented integration. The framework allows each algorithm to run independently, and realizes fast loading and online iteration of the algorithm.

In addition, from the perspective of the entire cloud system, software-defined cameras are an extension of the video cloud architecture. Cloud nodes located at the edge perception layer, cloud and intelligent capabilities can interact in the system. Through the transformation of the underlying technical architecture, traditional cameras can be realized. upgrade. For example, according to different application scenarios in different periods, different algorithms can be loaded on demand by time-sharing to realize rich intelligent applications such as intelligent identification, density statistics, behavior analysis, and structural analysis of people and vehicles in one camera. In addition, in response to the problem of light pollution, Ziguang Huazhi software-defined camera adopts dual-phase exposure technology, and adopts different exposure strategies for faces and vehicles, so that the capture rate of vehicles and faces can be higher in the scene where people and vehicles are mixed at night. Through the AI fill light, the fill light can be turned off when there is no one, and the fill light can be turned on after detecting a person to prevent avoiding capture and light pollution.

Software-defined cameras promote intelligence and enter the burst period

Software as a service, under the premise that the current intelligent security hardware has developed by leaps and bounds, starting from software, adapting to different service requirements in different scenarios, and upgrading security software services to meet the intelligent needs of multi-service and multi-scenario, and promote more areas of intelligence make a leap. Based on diversified and multi-level intelligent requirements, Unisplendour Huazhi has successively launched basic intelligence, Pratt & Whitney and fully structured software-defined cameras to adapt to more scenarios, meet the new needs of intelligent applications in the industry, and strive to solve the current intelligent problems and challenges faced. For traditional smart camera function curing, the smart function cannot be changed after leaving the factory. Software-defined cameras can define intelligent functions as needed, set different scene-based intelligence at different time periods and different preset points, load different algorithms, and realize different intelligent analysis functions, digest customers’ personalized/scenario-based algorithm demands, and reduce diversity. In addition, during the algorithm loading and unloading process, the live camera and recording services will not be interrupted, and the recording will not be lost; and it supports switching intelligent functions by time period and scene.

For example, in the education industry, for areas such as gates, gymnasiums, and other important activity entrances, crowd density detection algorithms are used during peak flow of people to achieve population density analysis and early warning in key areas to prevent dangers from occurring. In off-peak hours, the portrait algorithm can be switched to realize blacklist control. In addition, for the problem that the traditional camera intelligent algorithm cannot evolve, the intelligent algorithm of the software-defined camera can be quickly upgraded online (second level) without restarting the camera, and does not affect the audio and video Business, the live picture can be monitored normally, and the video recording is not interrupted. Through the continuous iteration of algorithms and applications, the intelligent effect is continuously enhanced, so that the camera has a longer lasting vitality.

The current wave of digitalization is turbulent, and the smart upgrade in the industry segment centered on the smart city has become unstoppable. As the "pioneer" of innovation and application in the field of "AI vision", software-defined cameras have become a cultural, educational and health, smart policing, smart It is a key demand point in many fields such as community and smart transportation. As a "sharp weapon" for Unisplendour Huazhi to empower the intelligent upgrade of all industries, software-defined cameras have been tested in practical applications in many provinces in China and in different fields, and will continue to strive to bring AI visual scene-based applications to the ground. Enter the second half of the intelligent upgrade of the industry.