Integrating real-time data analysis into automatic tracking of social insects
Alessio Sclocco1,2, Shirlyn Jia Yun Ong1, Sai Yan Pyay Aung1
1School of Biological Sciences, Nanyang Technological University, Singapore.
Royal Society Open Science
|May 7, 2021
Summary
We developed BACH, a real-time insect behavior analysis software using computer vision. While it tracks ant shapes well, individual identification struggles with fast-moving ants due to blurry matrix codes.
Area of Science:
- Ethology
- Computer Vision
- Bio-robotics
Background:
- Automatic video tracking is crucial for insect social behavior studies.
- Current systems often use offline analysis and computationally intensive methods.
- There's a need for real-time behavioral analysis tools.
Purpose of the Study:
- To develop BACH (Behaviour Analysis maCHine), a real-time video tracking software for insect groups.
- To evaluate BACH's performance against human observers in tracking ants.
- To assess the impact of ant speed on identification accuracy.
Main Methods:
- BACH utilizes convolutional neural networks for object recognition.
- Individual insects are identified using a matrix code recognition algorithm.
- Performance was compared to human observers using ant videos in a 2D arena.
Main Results:
- BACH detected ant shapes with performance close to human observers.
- Individual ant identification accuracy decreased significantly at higher speeds.
- Low efficiency in detecting matrix codes in blurry, fast-moving ant images was observed.
Conclusions:
- BACH demonstrates the potential for real-time data analysis in animal behavior research.
- Hardware and software adjustments are needed to improve BACH's limitations.
- Real-time analysis accelerates data generation and enables remote collaborative experiments.
