Register now After registration you will be able to apply for this opportunity online.
This opportunity is not published. No applications will be accepted.
Tracking Snowflakes with a Drone
In this project, we wish to characterize the size, shape, spatial distribution, and velocity of snow particles in the atmosphere from the vantage point of a multi-copter UAV, using computer vision techniques.
Snow precipitation is crucial for both local weather forecasting and global climate models. Surprisingly, we lack a basic understanding of how snowflakes interact within turbulent air, which affects their collision rate. How do they distribute in space? How often do they collide? Do they stick together? How fast do they fall?
In this project, we wish to characterize the size, shape, spatial distribution, and velocity of snow particles in the atmosphere from the vantage point of a multi-copter UAV, using computer vision techniques. The baseline setup will involve planar imaging with one camera, but more advanced and exciting would be a multi-camera system. The system will require software compensation for drone motion and an empirical study on image resolution vs illumination source and intensity. In collaboration with the Experimental Fluid Dynamics Lab of Professor Filippo Coletti, particle image velocimetry (PIV) will be used to further characterize the potential influence of UAV prop wash on the desired imaging air volume (and particles within). Once system parameters have been determined, field experiments are to be performed on a snowy night in a remote location away from city light pollution.
Snow precipitation is crucial for both local weather forecasting and global climate models. Surprisingly, we lack a basic understanding of how snowflakes interact within turbulent air, which affects their collision rate. How do they distribute in space? How often do they collide? Do they stick together? How fast do they fall?
In this project, we wish to characterize the size, shape, spatial distribution, and velocity of snow particles in the atmosphere from the vantage point of a multi-copter UAV, using computer vision techniques. The baseline setup will involve planar imaging with one camera, but more advanced and exciting would be a multi-camera system. The system will require software compensation for drone motion and an empirical study on image resolution vs illumination source and intensity. In collaboration with the Experimental Fluid Dynamics Lab of Professor Filippo Coletti, particle image velocimetry (PIV) will be used to further characterize the potential influence of UAV prop wash on the desired imaging air volume (and particles within). Once system parameters have been determined, field experiments are to be performed on a snowy night in a remote location away from city light pollution.
- Hardware (UAV & Camera)
- Define criterias: IP rating, payload, flight-time, mountable cameras etc.
- Evaluate different drones/cameras (e.g. DJI M300 / M600, Alta8, Zenmuse H20, Zenmuse XT2 etc.)
- Selection, ordering, setup & familiarization
- Software (Algorithm)
- Implementation of simple particle tracking algorithm
- Monocular particle tracking (first in vicon with fixed/moving target + fixed/moving handheld camera or drone)
- Experiments, evaluation, write-up
- Hardware (UAV & Camera) - Define criterias: IP rating, payload, flight-time, mountable cameras etc. - Evaluate different drones/cameras (e.g. DJI M300 / M600, Alta8, Zenmuse H20, Zenmuse XT2 etc.) - Selection, ordering, setup & familiarization - Software (Algorithm) - Implementation of simple particle tracking algorithm - Monocular particle tracking (first in vicon with fixed/moving target + fixed/moving handheld camera or drone) - Experiments, evaluation, write-up
- Drone enthusiast, experience with hardware
- Courses in computer vision, 3D vision, or geometric computer vision
- Hands-on experience in computer vision projects (e.g. stereo vision) is a plus
- Coding experience in C++ or Python is required
- Drone enthusiast, experience with hardware - Courses in computer vision, 3D vision, or geometric computer vision - Hands-on experience in computer vision projects (e.g. stereo vision) is a plus - Coding experience in C++ or Python is required