STL: Online Detection of Taxi Trajectory Anomaly based on Spatial-Temporal Laws

This project implements some taxi trajectory anomaly detection methods, including iBAT, iBOAT, OnATrade and my proposed STL.

Getting started

The trajectory anomaly detection consists of two part: trajectory pre-processing and online detection. Trajectory pre-processing converts the raw trajectory records into structured data. Online detection uses the processed data to detect the incoming trajectory.

In this project, the raw trajectory record is stored in Baidu Netdisk (code:w2uu), more detail of the which can be found in Dataset.

Dataset

• sh_taxi_data
  • Collected from Shanghai, China during Apr., 2015.
  • Field description: taxi ID, alarm, empty, ceiling light status, ?, brake, receive time, GPS time, longitude, latitude, speed, direction, #satellites
• sz_taxi_data
  • Collected from Shenzhen, Guangdong, China during Sep., 2009.
  • Field description: taxi ID, time, longitude, latitude, speed, direction, occupied
• sh_taxi_data
  • Collected from Chengdu, Sichuan, China during Aug., 2014.
  • Field description: taxi ID, latitude, longitude, occupied, time

Demo

STL Demo