Skip to content

Commit b023d84

Browse files
Combined Metro AI Applications under one architecture (open-edge-platform#256)
Co-authored-by: Jakub Sikorski <[email protected]>
1 parent d597f4b commit b023d84

File tree

278 files changed

+101292
-0
lines changed

Some content is hidden

Large Commits have some content hidden by default. Use the searchbox below for content that may be hidden.

278 files changed

+101292
-0
lines changed
Lines changed: 14 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,14 @@
1+
# This file is used to set environment variables for the smart transportation application.
2+
# Updating it manually is not recommended.
3+
# Instead, use the install.sh <APP_NAME> script to set it up automatically.
4+
# HOST_IP is setup automatically from the system.
5+
# To manually setup this IP address, run:
6+
# install.sh <APP_NAME> [HOST_IP] e.g. $ install.sh smart-parking 10.10.10.10
7+
8+
# This is the IP address of the host machine where the application will run.
9+
HOST_IP=0.0.0.0
10+
11+
# This variable identifies the name of the application to be used.
12+
# It should match the name of the application directory in the metro-ai-suite.
13+
# Supported applications: "smart-parking", "loitering-detection", "smart-intersection", "smart-tolling"
14+
SAMPLE_APP=loitering-detection
Lines changed: 111 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,111 @@
1+
# Metro Vision AI App Recipe
2+
3+
## Overview
4+
5+
The Metro Vision AI App Recipe suite provides an integrated AI-powered solution for modern urban traffic management and monitoring. This comprehensive platform combines multiple transportation-focused applications built on a unified architecture, enabling cities and organizations to deploy intelligent transportation systems efficiently.
6+
7+
**Available Applications:**
8+
- **Smart Intersection Management** (includes optional Scenescape components) - AI-driven traffic flow optimization and intersection monitoring
9+
- **Loitering Detection** - Real-time detection of loitering behavior in transportation hubs
10+
- **Smart Parking** - Automated parking space monitoring and management
11+
- **Smart Tolling** - Intelligent toll collection and vehicle classification
12+
13+
### Key Features
14+
15+
- **Real-time Video Analytics**: Process live video streams from traffic cameras with low latency
16+
- **Edge AI Processing**: Run inference locally to reduce bandwidth and improve response times
17+
- **Scalable Architecture**: Easily scale from single intersections to city-wide deployments
18+
- **Configurable Pipelines**: Customize detection parameters and object types for specific use cases
19+
- **MQTT Integration**: Seamless communication between components for event-driven responses
20+
- **Dashboard Monitoring**: Real-time visualization of traffic metrics and system performance
21+
- **Flexible Deployment**: Support for various camera types and video sources
22+
- **No-Code Configuration**: Set up complex rules and workflows through intuitive interfaces
23+
24+
25+
## How It Works
26+
27+
The Metro Vision AI App Recipe architecture is designed to facilitate seamless integration and operation of various components involved in AI-driven video analytics for transportation and safety & security scenarios. The system processes real-time video feeds from traffic cameras, applies deep learning models for object detection and classification, and generates actionable insights for traffic management.
28+
29+
![Architecture Diagram](docs/user-guide/_images/metro_app_arch.png)
30+
31+
### Core Workflow
32+
33+
1. **Video Ingestion**: Live camera feeds or video files are ingested into the DL Streamer Pipeline Server
34+
2. **AI Processing**: Deep learning models perform real-time inference for vehicle detection, pedestrian tracking, and behavior analysis
35+
3. **Metadata Extraction**: Rich metadata including object coordinates, classifications, and tracking information is generated
36+
4. **Event Processing**: Node-RED processes the metadata and triggers appropriate responses based on configured rules
37+
5. **Data Distribution**: Results are distributed via MQTT to dashboards, alerts, and external systems
38+
6. **Visualization**: Real-time video streams with overlaid analytics are displayed through WebRTC viewers
39+
40+
### Components
41+
42+
- **DL Streamer Pipeline Server (VA Pipeline):** The core video analytics engine that processes video frames, performs AI inference using optimized deep learning models, and extracts rich metadata including vehicle classifications, pedestrian detection, and traffic flow metrics.
43+
44+
- **Mosquitto MQTT Broker:** A lightweight message broker that facilitates real-time communication between all system components, enabling event-driven responses and seamless data flow throughout the smart city monitoring ecosystem.
45+
46+
- **Node-RED:** A visual programming platform that allows users to create custom traffic management rules, automated responses to traffic events, and integration workflows without requiring extensive coding knowledge.
47+
48+
- **WebRTC Stream Viewer:** Provides real-time visualization of processed video streams with AI-generated overlays, including bounding boxes around detected vehicles, pedestrians, and other objects of interest.
49+
50+
- **Grafana Dashboard:** An advanced monitoring and analytics platform that displays real-time traffic metrics, system performance data, historical trends, and customizable alerts for operator.
51+
52+
- **Inputs (Video Sources):** Supports various input types including IP cameras, RTSP streams, and video files, providing flexibility for different deployment scenarios and existing infrastructure integration.
53+
54+
#### Optional Components
55+
56+
- **Scenescape Management UI:** A web-based user interface for advanced scene configuration, camera calibration, and visual rule setup. Provides intuitive tools for defining detection zones, traffic lanes, and monitoring areas through a graphical interface.
57+
58+
- **Scenescape Controller:** The backend service that manages scene configurations, processes spatial analytics, and coordinates between the Management UI and the video analytics pipeline. Handles complex scene understanding and geometric transformations.
59+
60+
- **Scenescape Database (PostgreSQL):** A robust relational database that stores scene configurations, camera metadata, calibration parameters, and historical analytics data. Ensures data persistence and enables complex queries for reporting and analysis.
61+
62+
- **InfluxDB:** A time-series database optimized for storing and querying high-frequency transportation metrics such as vehicle counts, traffic flow rates, speed measurements, and system performance data. Enables efficient historical analysis and trend monitoring.
63+
64+
### Transportation-Specific Features
65+
66+
- **Vehicle Detection & Classification:** Accurately identifies and classifies different vehicle types
67+
- **Pedestrian Tracking:** Real-time tracking of pedestrians afor safety monitoring
68+
- **Incident Detection:** Automatic detection of accidents, stopped vehicles, or unusual traffic patterns
69+
- **Zone-Based Monitoring:** Configurable regions of interest for specific area monitoring
70+
- **Multi-Lane Support:** Simultaneous monitoring of multiple traffic lanes and directions
71+
72+
The DL Streamer Pipeline Server serves as the cornerstone of the solution, specifically optimized for transportation and traffic management scenarios. It leverages Intel-optimized deep learning models including pedestrian and vehicle detectors, traffic flow analyzers, and behavior recognition models to perform real-time inference on video streams.
73+
74+
The system is designed with professionals in mind, offering pre-configured pipelines for common traffic monitoring scenarios while maintaining the flexibility to customize detection parameters, tracking algorithms, and alert thresholds. This approach enables rapid deployment of AI-powered traffic management solutions without requiring deep learning expertise from traffic engineers and city planners.
75+
76+
Key advantages include edge-based processing that reduces latency for time-critical traffic decisions, support for multiple video sources to cover complex intersections, and integration capabilities with existing traffic management systems through standard protocols and APIs.
77+
78+
**Advanced Capabilities with Optional Components:**
79+
When deployed with the optional Scenescape components, the system provides enhanced spatial analytics capabilities including precise vehicle trajectory tracking, accurate speed measurement, and advanced zone-based analytics. The InfluxDB integration enables long-term trend analysis, capacity planning, and performance optimization based on historical traffic patterns.
80+
81+
## Use Cases
82+
83+
### Smart Intersection Management
84+
- Real-time traffic light optimization based on vehicle and pedestrian counts
85+
- Emergency vehicle detection and priority signal control
86+
- Pedestrian safety monitoring at crosswalks
87+
- Traffic violation detection and automated enforcement
88+
89+
### Loitering Detection
90+
- Security monitoring at transit stations and stops
91+
- Identification of suspicious behavior in hubs
92+
- Crowd management in high-traffic areas
93+
- Automated alerts for security personnel
94+
95+
### Smart Parking
96+
- Real-time parking space availability tracking
97+
- Automated payment and access control
98+
- Parking violation detection
99+
- Optimization of parking space utilization
100+
101+
### Smart Tolling
102+
- Automated vehicle classification for toll calculation
103+
- License plate recognition for electronic toll collection
104+
- Traffic flow optimization at toll plazas
105+
- Integration with existing tolling infrastructure
106+
107+
## To Get Started
108+
- [Smart Intersection](smart-intersection/docs/user-guide/Overview.md)
109+
- [Smart Tolling](smart-intersection/docs/user-guide/Overview.md)
110+
- [Smart Parking](smart-parking/docs/user-guide/Overview.md)
111+
- [Loitering Detection](loitering-detection/docs/user-guide/Overview.md)

0 commit comments

Comments
 (0)