FYI: We use OpenCV only for video input and visualization. Most of the algorithms are implemented using C/C++. I am improving the readability of this code. Will update it soon!

Abandoned Object Detection via Temporal Consistency Modeling and Back-Tracing Verification for Visual Surveillance

Created by Kevin Lin, Shen-Chi Chen, Chu-Song Chen, Daw-Tung Lin, Yi-Ping Hung at National Taiwan University.

This paper presents an effective approach for detecting abandoned luggage in surveillance videos. We combine short- and long-term background models to extract foreground objects, where each pixel in an input image is classified as a 2-bit code. Subsequently, we introduce a framework to identify static foreground regions based on the temporal transition of code patterns, and to determine whether the candidate regions contain abandoned objects by analyzing the back-traced trajectories of luggage owners. Experimental results show that our method performs more favorable against the others.

The details can be found in the following IEEE TIFS 2015 paper

If you find our works useful in your research, please consider citing:

Abandoned Object Detection via Temporal Consistency Modeling and Back-Tracing Verification for Visual Surveillance
K. Lin, S.-C. Chen, C.-S. Chen, D.-T. Lin, and Y.-P. Hung
IEEE Transactions on Information Forensic and Security (TIFS), 2015  


Left-Luggage Detection from Finite-State-Machine Analysis in Static-Camera Videos
K. Lin, S.-C. Chen, C.-S. Chen, D.-T. Lin, and Y.-P. Hung
International Conference on Pattern Recognition (ICPR), 2014 

0. OpenCV

Simply run the following commands:

$ cmake .
$ make
$ ./download_video.sh

This demo detect abandoned luggage in the video

$ ./ObjLeft

Select the input source (1: video, 2: camera)

$ 1

Input the filename if you choose video

$ pets2006_1.avi

Click a rectangular Region of Interest (ROI) for detection

Double-press any key and start detection

This demo will detect the abandoned luggage and display the owner trajectory.

Please feel free to leave suggestions or comments to Kevin Lin ([email protected])