Smart Parking provides information on which parking spaces are occupied based on video feeds from multiple cameras, each monitoring a specific section of a parking lot. It is an end-to-end solution offering an analytics dashboard with both real-time and historical results.
It is a cloud-native application composed of microservices, using pre-trained deep learning models for video analysis. The app is modular and highly configurable, enabling you to adjust its parameters, such as detection thresholds and object types, without the need for extensive coding or deep learning expertise. Its key features are:
- Vision Analytics Pipeline: detects and classifies objects using pre-configured AI models.
- Integration with MQTT, Node-RED, and Grafana: facilitates efficient message handling, real-time monitoring, and insightful data visualization.
- User-Friendly: prebuilt scripts and configuration files facilitate ease of use.
- Get Started
- System Requirements
- Deploy with Docker Compose
- Deploy with Helm
- Deploy with Edge Orchestrator
Here is an overview of the architecture and logic of the application.
- DL Streamer Pipeline Server is a core component of the app. It receives video feed from multiple cameras (four by default, simulated with a video recording). With pre-trained deep learning models, it performs real-time object detection, classification, and tracking. It recognizes vehicles in the parking lot and sends their 2D bounding boxes to Node-Red, through the MQTT Broker. It also adds the detected bounding boxes on top of the video input, consumed by the WebRTC Server.
- Mosquitto MQTT Broker is a message distribution service, passing data between these sends the raw coordinates of detected vehicles to Node-RED. The feedback it receives is moved to Grafana to display.
- Node-RED, is a low-code platform for setting up application-specific rules and triggering MQTT-based events. It takes the 2D bounding boxes of detected cars and matches them with predefined parking slot positions, thus determining if a slot is free or occupied.
- Grafana Dashboard provides a monitoring and visualization dashboard indicating which and how many parking spaces are occupied. It also displays the original video feed coming from the WebRTC Server.
- WebRTC Server serves video streams processed by the pipeline for end-user visualization. It is supplemented by the Coturn signaling server and passes the feed for display in Grafana.
- Inputs (Video Files and Cameras): Provide raw video streams or files as input data for processing in the pipeline.
- Nginx: is a high-performance web server and reverse proxy that provides TLS termination and unified HTTPS access.