Augmented reality application for AirPort Security

Challenge

The real time video processing requires a lot of the computation power. It is not possible to put so efficient CPU in the googles, due to limited size of googles and battery efficiency. So the most important challenge was to design the architecture of the complete solutions, which will provide enough computation power needed for video analysis and face recognition.

The second challenge was to design solution which works near real time. The terrorists and suspected persons need to be identified in near real time to allow security services to take proper actions. Even small delay with face identification could cause that suspected persons can perform dangerous actions.

The third challenge was to minimize cost of the complete system. As the complete solution is very complex and requires a lot of computational power, the very important aspect is to minimize the cost of the system.

Solution

Holisun has chosen the fully cloud based solution with processing video in the cloud. Thanks to using MELODIC/Functionizer platform and deploy video processing and facial recognition components using serverless model, the complete solution fulfills the requirements. The serverless components have the ability to start operate very fast, in the matter of single seconds comparing to minutes needed to start virtual machines. Also the cost is much lower, especially comparing to VMs running continuously. Also the serverless components are very easy to scale, it is possible to start up to 3000 instances simultaneously.  The application deployment is fully managed by MELODIC platform, which allows for optimizing deployment both for performance and cost purposes. MELODIC controls the need for the components instances based on the real time monitoring of the application’s performance. In case of delay in face recognition tasks, MELODIC automatically added new instances of the needed components.

Benefits

Using serverless components optimized by MELODIC allows for up to 60% of the cost savings. The same solution deployed using traditional continuously working VMs would be significantly more expensive, than serverless approach. The second benefit is very fast and dynamic scaling of the solutions. In case of needs for more performance, the new instances of the components can be started in seconds instead of minutes. And the number of processing threads can be much higher than using traditional approach with virtual machines.

The presented solution set new standard for the real time video processing and enable the usage of augmented reality applications. Especially real time back-end video stream processing allow for almost unlimited number of the use cases in various domains of life.