Comparative analysis of tracking and behavioral patterns between wild-type and genetically modified fruit flies using...

Area of Science:

• Collective Animal Behavior
• Biophysics
• Robotics and Automation

Background:

• Collective animal behavior is observed across diverse species but is challenging to quantify due to data acquisition difficulties.
• Existing models can qualitatively represent group behaviors, yet precise data from real organisms is crucial for validation.
• Tracking individual movements within swarms is complex, especially with sudden changes in speed or direction.

Purpose of the Study:

• To introduce a novel tracking algorithm, the Hidden Markov Unscented Tracker (HMUT), for precise analysis of collective animal behavior.
• To capture and analyze 3D flight trajectories of fruit fly swarms.
• To investigate fruit fly behavioral characteristics using statistical and clustering methods on trajectory data.

Main Methods:

• Developed the Hidden Markov Unscented Tracker (HMUT) by integrating Hidden Markov Models (HMM) for prediction and Unscented Kalman Filters (UKF) for nonlinear processing.
• Captured fruit fly swarm movement using stereo cameras and detected individual fruit flies with the AKAZE algorithm.
• Reconstructed 3D trajectories using polar coordinate constraints and analyzed motion patterns using Dynamic Time Warping (DTW) for trajectory similarity.

Main Results:

• The HMUT algorithm demonstrated efficient and accurate tracking of fruit fly swarms, effectively handling changes in motion.
• Generated high-resolution 3D motion data of individual fruit flies within a swarm.
• Identified distinct behavioral characteristics and quantified similarities and differences within fruit fly clusters based on flight trajectories.

Conclusions:

• The HMUT algorithm provides a robust framework for tracking and analyzing complex collective animal behaviors.
• The study successfully generated detailed 3D flight data, offering new insights into fruit fly swarm dynamics.
• This approach facilitates a deeper statistical and similarity-based understanding of group movement patterns in biological systems.