Trajectory Recognition
Subtracting the static background of videos is a powerful technique used to isolate moving objects. In our case, birds and insects are being filmed passing through the horizon. We subtract this static horizon with a carefully designed software that uses background subtraction. The software automatically tracks the trajectories of the birds and insects through clustering algorithms and combines these trajectories into a single picture. Using these trajectories, the varying flight patterns allow indications for differentiating birds from insects.
Functional Specifications
The starting point for the trajectory recognition project is a camera that is positioned on the roof of the physics building in Frankfurt in a fixed and static set up. The camera films the roof of the building and the horizon as the scenery. There are different objects like birds and insects passing through the scenery. The main goal is the separation of the trajectories of the birds and insects from the background. The fixed position of the camera formed a sound basis for efficient background subtraction. Airplanes, spyder webs, multiple objects flying through the background, reflections of the sun on the roof, and different weather conditions are the main problems for a successful separation. Different algorithms have been tested to find the best result.
Example
Availability
Description
A more detailed description is available in the GitLab project wiki.
Source Code
The source code is released under the MIT License on GitLab: https://gitlab.com/ganymede/trajectory-recognition
Presentation
The accompanying poster, presented on , is available under CC-BY 4.0 license on GitLab: https://gitlab.com/ganymede/trajectory-poster
