The goal of this project is to develop an automatic real time incident detection System for Hong Kong. I.e. an application which tells you whether any given moment in any given road there is an accident. This project is very similiar to what Google does, however in addition to GPS data I have included Road Sensor data into our analysis. A quick intro to this project via a poster
Accurate incident detection is important mostly for 2 reasons: Routing and Acknowledging. For routing algorithms, incident detection knowledge gives ground to a better estimate of travel times for different routes. Automatic incident detection can as well be useful for informing important agents like the police or traffic authority about the incident so they could take necessary action faster.
This webpage divides into 3 parts:
- Data preprocessing development of an universal system to parse GPS, accident and road sensor (speed panel) info into a form which can be inserted to different incident detection models/algorithms.
- LDA applicaton - model analysis - The traffic accident detection is based on this paper: [IS’15] Real-time Traffic Incident Detection Using a Probabilistic Topic Model . Our goal is to improve that method by including road sensor data into the model and conducting theoretical analysis.
- Result Analysis - Experimental analysis of the incident detection. This analysis seeks to be the baseline for comparing different incident detection methods on Hong Kong data.
All work in this project is supervised by