ccu
60
edits
Changes
Jump to navigation
Jump to search
Created page with "== Overview == The ability of swarms to form structures was shown to be present in small insects like ants. The Eciton army ant form collective structures from them itself t..."
== Overview ==
The ability of swarms to form structures was shown to be present in small insects like ants. The Eciton army ant form collective structures from them itself to support the colony. This project aims to measure the activity of individuals on their tail to collect food through a plane with different adjustment of tilt to encourage the swarm to form a supporting structure which leads higher efficiency of the whole colony in transporting food.
The experiments were performed with a static plane sensed with a conventional camera under in the outdoor setting where the illumination is dynamically changing.
== Contact ==
* [[[https://www.cvia.uni-konstanz.de/members/prof-dr-bastian-goldluecke/ Prof. Dr. Bastian Goldluecke]]]
* [[[http://collectivebehaviour.com/people/lutz-matthew/ Matthew Lutz]]]
* [[https://www.cvia.uni-konstanz.de/members/research-staff/msc-antonin-sulc/ Antonin Sulc]]]
== Aims ==
Track the individual ants and count their activity
== Estimated level of difficulty ==
There are several challenges related to tracking individual ants:
* The individuals are visually very similar
* The individuals get spatially very close over time
* The scenes are captured in the wild thus illumination is changing over time
The first two problems cause that the classical model based trackers cannot be used because= they are usually designed to objects which have at least slightly distinct colors. b
== Provided data ==
== Suggested/tested approaches ==
The approach was unsuccessfully tested with the classical model based trackers (Boost, MIL, MedianFlow). The model-based trackers need at least one positive sample of the object in order to build a model to track. For the above-mentioned problems, the model-based trackers failed on this task.
The setting is, however, except lighting conditions, relatively static because other non-tracked parts remain fixed thus we can use the background subtraction which provides unreliable, but sufficiently accurate detections which can be passively tracked with the multi-object tracking approaches like [1,2]
* [1] [[http://vision.cse.psu.edu/courses/Tracking/vlpr12/lzhang_cvpr08global.pdf Global Data Association for Multi-Object TrackingUsing Network Flows]]
* [2] [[http://me8xh5ls2s.scholar.serialssolutions.com/?sid=google&auinit=H&aulast=Pirsiavash&atitle=Globally-optimal+greedy+algorithms+for+tracking+a+variable+number+of+objects&id=doi:10.1109/CVPR.2011.5995604 Globally-optimal greedy algorithms for tracking a variable number of objects]]
The ability of swarms to form structures was shown to be present in small insects like ants. The Eciton army ant form collective structures from them itself to support the colony. This project aims to measure the activity of individuals on their tail to collect food through a plane with different adjustment of tilt to encourage the swarm to form a supporting structure which leads higher efficiency of the whole colony in transporting food.
The experiments were performed with a static plane sensed with a conventional camera under in the outdoor setting where the illumination is dynamically changing.
== Contact ==
* [[[https://www.cvia.uni-konstanz.de/members/prof-dr-bastian-goldluecke/ Prof. Dr. Bastian Goldluecke]]]
* [[[http://collectivebehaviour.com/people/lutz-matthew/ Matthew Lutz]]]
* [[https://www.cvia.uni-konstanz.de/members/research-staff/msc-antonin-sulc/ Antonin Sulc]]]
== Aims ==
Track the individual ants and count their activity
== Estimated level of difficulty ==
There are several challenges related to tracking individual ants:
* The individuals are visually very similar
* The individuals get spatially very close over time
* The scenes are captured in the wild thus illumination is changing over time
The first two problems cause that the classical model based trackers cannot be used because= they are usually designed to objects which have at least slightly distinct colors. b
== Provided data ==
== Suggested/tested approaches ==
The approach was unsuccessfully tested with the classical model based trackers (Boost, MIL, MedianFlow). The model-based trackers need at least one positive sample of the object in order to build a model to track. For the above-mentioned problems, the model-based trackers failed on this task.
The setting is, however, except lighting conditions, relatively static because other non-tracked parts remain fixed thus we can use the background subtraction which provides unreliable, but sufficiently accurate detections which can be passively tracked with the multi-object tracking approaches like [1,2]
* [1] [[http://vision.cse.psu.edu/courses/Tracking/vlpr12/lzhang_cvpr08global.pdf Global Data Association for Multi-Object TrackingUsing Network Flows]]
* [2] [[http://me8xh5ls2s.scholar.serialssolutions.com/?sid=google&auinit=H&aulast=Pirsiavash&atitle=Globally-optimal+greedy+algorithms+for+tracking+a+variable+number+of+objects&id=doi:10.1109/CVPR.2011.5995604 Globally-optimal greedy algorithms for tracking a variable number of objects]]
