Intel OpenVINO™ Smart Intersection Analytics

Smart Intersection is an AI-powered edge video analytics solution designed for intelligent transportation and smart city scenarios. It combines feeds from multiple cameras to track vehicles from different angles, analyze their speed and direction, and understand interactions in real space. You can further integrate it with real-time traffic monitoring, object detection, pedestrian tracking, and intersection analytics using Intel optimized AI pipelines.

How It Works

This section illustrates how the application integrates with the edge AI system architecture, including video ingestion, AI inference, analytics, and visualization components.

Upstream Reference

This work is based on the official Intel Metro AI Suite reference implementation:

• Intel Smart Intersection

Prerequisites

• Verify that your system meets the minimum requirements.
• Ensure Git.
• Install Docker Engine.
• Perform post-installation steps - configure Docker as non-root user.
• Utilizing Docker Compose for streamlined deployment.
• OpenVINO Device Support:
  • Integrated GPU (iGPU)
  • Neural Processing Unit (NPU)

Deploy Smart Intersection

1. Clone Repository

git clone git@github.com:Advantech-Containers/Smart-City-Intel-Metro-AI-Smart-Intersection-Application-powered-by-Intel-OpenVINO.git
cd Smart-City-Intel-Metro-AI-Smart-Intersection-Application-powered-by-Intel-OpenVINO 

2. Install Assets

./install.sh smart-intersection

3. Start Application

export SUPASS=$(cat ./smart-intersection/src/secrets/supass)

docker compose up -d

After that, you can verify running containers with docker ps.

Expected services include:

• Grafana
• DL Streamer Pipeline Server
• MQTT Broker
• Node-RED
• SceneScape Services
• InfluxDB

Configure Inference Device (CPU / GPU / NPU)

By default, the Smart Intersection pipelines run on CPU inference.
This application also supports Intel GPU and NPU acceleration through OpenVINO optimized pipelines.

Inference device selection is configured through:

smart-intersection/src/dlstreamer-pipeline-server/config.json

For detailed configuration steps, please refer to:

• How to use GPU for inference
• How to use NPU for inference

Access Web Interfaces

Application UI

• Log in with credentials:
  • Username: admin
  • Password: Get it by echo $SUPASS

Once the app is up, open the browser, navigate to https://localhost. Input the Username and Password as mentioned.

• Intel SceneScape Sign-in UI

Upon successful login, you will be redirected to the landing page.

To verifying the application:

1. Navigate to the Scene: In the Scene Management UI, click the Intersection-Demo card.

2. Observe Fused Tracks: On the Scene page, confirm that fused object tracks are moving on the map.

3. View Camera Feeds: Check the camera frames (initially greyed out). Toggle the Live View button to enable incoming streams.

4. Confirm Correlation: Ensure that object detections in the camera feeds align with the tracks displayed on the map.

5. Grafana Dashboard:

• Navigate: Open the Grafana UI and go to the home page. Log in with credentials:

  • Username: admin
  • Password: admin (You will be prompted to change it on first time)

• Access Dashboard: Click on Dashboards and select Anthem-ITS-Data.

• Observe: View aggregated analytics for various Regions of Interest (ROIs).

Stop the Application

To stop and remove all running microservices, execute the following command:

docker compose down -v

References

• Intel Open Edge Platform

• Metro AI Suite

• Smart Intersection