Achieve business utilization of video AI! What is the strength of EdgeMatrix based on WEBRTC?

The cloud type provided as a service is easy to install and set, so it is easy to try small projects.If the camera can be connected to a cloud service, it can be easily used, and the storage capacity can be expanded flexibly.

Although it is a cloud type that is being introduced, there are several issues.Storage and computing are often affected by services and maintenance because they depend on the cloud.Many of the functions are provided by vendors and cannot be completely controlled.In some cases, it is not desirable to store video in the cloud due to security and privacy issues, such as the face and appearance of a person.

"The biggest problem is the network. It is not suitable for real -time detection or analysis because it sends images to the cloud and processed. For example, when used for crime prevention, we analyze images and lightly turn on lights.It is meaningless if you can not warn or not. If you use it in remote areas such as mountainous areas, there are cases where network performance can be a problem, "Sato pointed out.

What is attracting attention is the edge type that utilizes edge computing.Edge devices are installed near the camera where the video is generated, and AI processing can be performed in real time on site.Learning and managing is performed with cloud services, but inference continues to function even if the cloud stops.Since data transfer is unnecessary, there is an advantage that high analysis accuracy can be exerted with high -quality images.It is also possible to select a method that does not store images in the device itself, considering security and privacy.By transmitting only the processing results to the cloud side, the network efficiency is good, and it works well in low -band environments.

Solved three issues of video edge AI on Webrtc

However, there are also issues on edge -type video platforms, and there are three mainly three.

One is unauthorized access to edge devices.Since it is often installed outdoors, it is highly risk of attacking methods such as fraudulent devices because others can physically contact.How to protect the images and applications stored in the edge and how to identify the edge identity is a problem.

The second is how to browse the live video of the camera and the recorded video.Many edge devices are connected to the network using LTE / 5G lines.To access the private address of this device, there is a method of setting a port mapping with a router or connecting with VPN.The former cannot be set easily because the safety is greatly impaired.The latter needs to use VPN devices or services, and engineers who design and maintain networks are indispensable.Therefore, it is not suitable for use in remote areas and high places.

Finally, it's a commercial power supply.When installing a power supply, such as the back of the mountain, it is important to note whether it can support video AI, assuming that it can be used in places where it is difficult to secure a natural energy and storage battery.

"The EdgeMatrix video platform has adopted" WebRTC "technology as a core for solving these three issues. By built on a P2P network by WEBRTC, security and scalability issues are low -cost.I succeeded in solving it "(Mr. Sato)

Edgematrix Co., Ltd. Managing Executive Officer Product & Service Headquarters Platform Development Unit Takeshi Sato

Webbrtc is one of the most notable technologies and realizes real -time communication to web browsers and mobile apps via a simple API.Recently, it has been frequently used in Web conferencing applications, and since 2020, it has rapidly developed technical specifications.EDGEMATRIX video platform uses WEBRTC technology extensively, but there are two major points according to Sato.

One is that there is no need to introduce a special plug -in or device to connect private addresses.A typical web browser is available, and special equipment is not required on the edge side.The advantage is that it is easy to install in remote areas described above and that services can be provided to consumer users.

The second is that scalability can be secured.The p2P communication method that does not go through the server has the drawback that the client needs to perform a heavy encoding, so that the load on the media transmission is increased in proportion to the number of connections.Therefore, EdgemaTrix adopted a design that was equipped with a hardware encoder and shared this processing power with a client.This is a method that could be realized because it is a webrtc, the CPU is very low load and a large -scale service environment can be built.This characteristic is very important, especially in commercial use, because there are cases where the environment is opened to many users (allowing video viewing).In addition, it is suitable for edge -type video platforms in that power consumption can be reduced.

Grow EDGEMATRIX to various businesses

In 2020, the field of video AI / video platform has progressed remarkably.Above all, the progress of the camera is quickly, and AI has been installed on the market for the market.In other words, general -purpose AI inference can be performed on the camera side, so a video platform requires more advanced and specialized AI algorithms.

"EdgeMaTRIX supports" EMPYCUSTOM ", which tracks images in consideration of time series. In the past, it was actually treated and connected to the images that were cut out, so we recognize and analyze human movements.It was difficult to do. If you use this technology, you can detect people's movements as continuous activities, so you can expect them to be used for various services. "(Mr. Sato).

EdgeMatrix will continue to expand the functions of video platforms.One of them is a plan to take advantage of the features of P2P and sublimate it into a function that can be experienced specifically.

What we particularly expect is the use of business.It is expected that video platforms are not only used in the company closed in the company, but also provides a case to third parties, such as cases where local governments distribute video services to citizens.It is a method that was not considered in the conventional VMS + VPN format.

In addition, the use as a surveillance camera has been enhanced to work together with video viewing and display from various alerts.Various cooperation methods can be expected, such as using existing IoT sensors, receiving metadata emitted by AI -equipped cameras and using them as alerts.In the future, it may be possible to make some voice a trigger.

"We are also developing a function to connect to other Webbrtc clients, such as the Web Conference System. In other words, it is possible to access the edge images from Google Meet and Zoom, which uses the mobile apps that you always use as a viewer.You can expect various ways to use it, such as displaying video of the site during the meeting and consulting about how to respond. "(Mr. Sato).

EdgeMatrix wants to return to the VMS video platform, regardless of the edge.The advantages of existing EdgeMATRIX solutions can be performed with numerous applications with numerous cameras, but there is also a need to share the same application with many cameras, including nursing care facilities and commercial facilities.

They are also planning to collaborate with a very simple monitoring system and network video recorder (NVR).NVRs are easy to use, but they cannot be expected to cooperate with other systems like VMS.We are developing a mechanism to send analyzed streams to NVR.In addition, in order to cooperate with VMS more simply, the function of forwarding the time stamp and inference result as metadata and the function of forward to VMS is planned.

EDGEMATRIX, which summarizes the necessary devices, networks, and cloud infrastructure as a solution based on the video edge AI, will continue to evolve in the future.I would like to pay attention to the video platform that can perform advanced inference in real time in various fields and provide modern services.

[PR] Provided: EdgeMatrix