top of page
Jon Zhao

OpenNCC WE: PoE和WiFi使能边缘AI物联网摄像头感知世界



In the deployment scenario of edge AI cameras, we are faced with the requirements and challenges of installation distance, cable, power supply, wireless and Internet of Things integration. OpenNCC WE is designed to address these scenarios in intelligent deployment.

Various deployment possibilities:

  1. POE Power Supply Integration

  2. POE Power Supply with WIFI connection

  3. 12V DC with Ethernet cable

  4. 12V DC with WIFI connection

  5. Solar panels and battery packs with WIFI connection

  • In the deployment of large traditional network systems, how to find a new Edge AI cameras is very important, OpenNCC WE supports the ONVIF protocol to discover cameras.

  • If you want your camera to connect directly to the cloud, OpennCC WE reserves the ability to let you integrate with AWS IOT.

  • Deploy your own model remotely,compatible with OpenVINO.

How it works

Power usage

The OpennCC WE accepts power input from the 802.3af. It can also accect power from 12V@1A DC power adapter.

Power usage for OpennCC WE ranges between 1.5 W (standby) and 7 W (max consumption with WIFI).


AI Model inference deployment

System framework and introduction

From a software system perspective, OpenNCC WE consists of the following development packages:

A generic OpenNCC development SDK running on MediaTek SoC on OpenNCC WE. Responsible for obtaining video stream from OpenNCC VPU camera through USB2.0, controlling camera, downloading AI model and obtaining AI-Meta stream.

2. OpenNCC IPC SDK for device side

Also run on MediaTek SoC side,this part of the source code is responsible for: RTSP protocol stack, ONVIF protocol, system watchdog, and TCP proprietary protocol framework.

The OpenNCC team has helped users develop reference designs for streaming media management, network camera framework, and AI inference management framework and open source sharing them.

3. OpenNCC IPC SDK for client side

This is part of the IPC SDK, but runs on the side that needs to manage and control OpenNCC WE, and get video streams and AI results from it. It can be PC, server, NVR and edge box.However, the default github version is based on PC, other platforms can contact FAE to obtain.

  • The client library under: ClientSdk

  • OpenNCC IPC Viewer is a QT based demo tool,which already integrated the ClientSdk,and could use it to display video, extract the AI results and draw box on the window, download AI-model to OpenNCC WE.

  • For systems integrators, just pay attention on OpenNCC IPC SDK for client side.

  • If you want to develop application on MediaTek SoC to integrate into your cloud or special systems, you could contact local FAE.

Start with OpenNCC IPC Viewer

1. Git clone the repo from github:

$ git clone https://github.com/EyecloudAi/openncc_ipc.git


2. Enter your openncc ipc sdk installed path,then:

$ cd ipc_viewer/bin/IpcViewer_1.0.0

$ sudo ./AppRun

Would pop up a window as following:

3. Prepare network connection

We need connect the camera to a POE switch,and power on the OpenNCC WE camera.

The computer that want to access this camera needs to be connected to the same LAN with the openncc we camera.


4. The PC and the OpenNCC WE camera are connected to a LAN


5. Or the PC could connect to OpenNCC WE camera's WIFI AP 'eyecloud_ipc'


6. Click 'Scan devices' on IPC Viewer demo,it would list out all the cameras under same LAN,which support ONVIF.


7. Select one of the scanned OpenNCC WE camera,then you could select one AI-model to download to the camera. 10 models already has been integrated and can be selected.


8. Click 'Update model',and click 'Play' to display the stream,and the demo application would extract the AI results and draw box on the image.


Start with OpenNCC IPC libaray

Under ClientSdk directory,you could see:

├── ClientSDK_API.pdf

├── CliTest

├── inc

├── libClientSdk.a

├── main.cpp

└── Makefile

  • You can refer to the API documentation,it introduces the function in detail.

  • main.cpp is the reference code.

  • The libClientSdk.a for X86 platform.

34 次查看0 則留言

Comentarios


bottom of page