Multiple Ant Tracking

Overview

Motion and behavior analysis of social insects such as ants requires tracking many ants over time. This process is highly labor-intensive and tedious. Automatic tracking is challenging as ants often interact with one another, resulting in frequent occlusions that cause drifts in tracking. In addition, tracking many objects is computationally expensive. In this paper, we present a robust and efficient method for tracking multiple ants. We first prevent drifts by maximizing the coverage of foreground pixels at at global scale. Secondly, we improve speed by reducing markov chain length through dynamically changing the target proposal distribution for perturbed ant selection. Using a real dataset with ground truth, we demonstrate that our algorithm was able to improve the accuracy by 15% (resulting in 98% tracking accuracy) and the speed by 76%.

Download

Data

The data is consisted a video of 49 Temnothorax rugatulus ants over 1000 frames, taken at thirty frames per second. The frames are extracted into 1000 individual PNG images which located in the antdata.zip. The naming convention of the images is FRMx where x is the 5 digit frame number (from 00001 to 01000).

Ground-truth

The ground-truth, antdataGT.zip, contains the manually annotated ants' position parameters and a Matlab loading script.
Each row in the text file colony05GT1KFrames.txt contains: AntNumber, FrameNumber, X-cooridinate, Y-coordinate, and Orientation.
The ground truth can be automatically loaded into a Matlab structure using the script loadGroundTruth.m. To load the ground truth, type the following into Matlab command prompt:

>> gtArray = loadGroundTruth('colony05GT1KFrames.txt');

where gtArray : an array (M X N) of structures where M is the number of ants in the video and N is that number of frames. Each structure contains 3 fields: xt is the x-coordinate position, yt is the y-coordinate position, and ta is the orientation with resect to the x-axis.

Code Bundles

The project is implemented in MATLAB. The program is free for non-commericial usage. We encourage other researchers to collaborate.
Download Code Bundles (Coming soon)

Publications

  1. Corey Poff, Hoan Nguyen, Tim Kang, Min Shin. "Efficient Tracking of Ants in Long Video with GPU and Interaction." IEEE Workshop on Applications of Computer Vision (WACV). Breckenridge, Colorado. January 2012. PDF WACV

  2. M. Fletcher, A. Dornhaus, M. C. Shin. "Multiple Ant Tracking with Global Foreground Maximization and Variable Target Proposal Distribution" IEEE Workshop on Applications of Computer Vision (WACV). Kona, Hawaii. January 2011. PDF WACV

People

Min Shin
Min Shin, PhDAssociate Professor in Computer ScienceComputer Science Department, UNC CharlottePhone: (704) 687-8578Office: 403D WoodwardEmail: mcshin at uncc dot edu
Rich Nguyen
Nhat 'Rich' Nguyen, MSPhD Student in Computer ScienceComputer Science Department, UNC CharlottePhone: (704) 687-8582 Office: 404 WoodwardEmail: nhnguye1 at uncc dot edu
Rich Nguyen
Corey PoffStudent in Computer ScienceDavidson CollegePhone: N/AOffice: Davidson, NCEmail: coreypoff at gmail dot com
Rich Nguyen
Mary FletcherStudent in Computer ScienceColby CollegePhone: N/A Office: Waterville, MEEmail: mvfletch at colby dot edu
Rich Nguyen
Anna Dornhaus, PhDAssociate Professor in BiologyDepartment of Ecology and Evolutionary Biology, University of ArizonaPhone: N/A Email: dornhaus at email dot arizona dot edu
Rich Nguyen
Thomas Fasciano, MSPhD Student in Computer ScienceComputer Science Department, UNC CharlottePhone: (704) 687-8582 Office: 404 WoodwardEmail: twfasciano at uncc dot edu
Rich Nguyen
Hoan Nguyen, BSMaster Student in Computer ScienceComputer Science Department, UNC CharlottePhone: (704) 687-8582 Office: 404 WoodwardEmail: hdnguye1 at uncc dot edu