This repository contains the script and a sample data for multi-vehicle tracking in a panoramic scene. The data was collected on July 23rd, 2016 in Tucsan, AZ. The test footage lasted for about 7 minutes and 22 seconds and recorded detailed movement of 20 vehicles. The script is used to extract and visualize the trajectories of those 20 vehicles in the scene.
The shell script is tested in Linux environment has the following dependencies:
- ffmpeg
- Python 2.7
- OpenCV Python binding
- Scipy stack
- scikit-learn
To run the video processing pipline,
>> mkdir az_0726_expB
>> wget https://www.dropbox.com/s/urcmdzqwdnbdkkg/original_video.mp4?dl=0
>> ./run.sh
This repository only contains data for one single experiment. However, a total of eight experiments were conducted in Tuscan in July 23rd. To access the full stack of video footages, please go to here.
The vehicle movement data generated using the script, combined with the simultaneous fuel rate measured from a OBD scanner, is publicly available here.
- Author: Fangyu Wu, Coordinated Science Laboratory, UIUC
- Email: fwu10(at)illinois.edu
- Web: fangyuwu.com