Register now After registration you will be able to apply for this opportunity online.
This opportunity is not published. No applications will be accepted.
Onboard Position Estimation for Intelligent Drone Cinematography
The focus of the thesis is to implement, evaluate and test a novel indoor positioning system for drone cinematography. The technology is based on infrared markers and uses visual odometry for localization. In a first step, the student should re-implement and test the system using offline data. In a second step, the estimation code should be implemented onboard on a Parrot Bebop 2 drone.
Smooth shots have been a crucial demand in the media and entertainment industry ever since its first days.
Creating spectacular images for sporting events or action movies, however, still requires heavy camera equipment such as dollies or cranes.
These methods are very expensive and inconvenient: Due to their size and weight, transportation is cost-intensive and the setup often requires two or more people and a lot of time. Additionally, there are limitations: For sports productions, available options offer hardly longer tracks than 20m and there are speed limits of about 40km/h.
Drones are famously known for their aerial videography capabilities. However, state-of-the-art drones face two major issues, preventing them from being used as a compelling replacement for dolly movements:
1. Filming often requires more than one take of the same shot. Therefore, drones need to be able to refly the exact same path several times. This is hardly possible with a manually controlled drone, and if so, it requires hours and hours of training.
2. Computer-controlled drones may solve this problem. However, their current positioning systems are not reliable enough yet, due to the lack of precision technology.
Smooth shots have been a crucial demand in the media and entertainment industry ever since its first days. Creating spectacular images for sporting events or action movies, however, still requires heavy camera equipment such as dollies or cranes. These methods are very expensive and inconvenient: Due to their size and weight, transportation is cost-intensive and the setup often requires two or more people and a lot of time. Additionally, there are limitations: For sports productions, available options offer hardly longer tracks than 20m and there are speed limits of about 40km/h.
Drones are famously known for their aerial videography capabilities. However, state-of-the-art drones face two major issues, preventing them from being used as a compelling replacement for dolly movements:
1. Filming often requires more than one take of the same shot. Therefore, drones need to be able to refly the exact same path several times. This is hardly possible with a manually controlled drone, and if so, it requires hours and hours of training. 2. Computer-controlled drones may solve this problem. However, their current positioning systems are not reliable enough yet, due to the lack of precision technology.
The focus of the thesis is to implement, evaluate and test a novel indoor positioning system for drone cinematography. The technology is based on infrared markers and uses visual odometry for localization. In a first step, the student should re-implement and test the system using offline data. In a second step, the estimation code should be implemented onboard on a Parrot Bebop 2 drone.
The focus of the thesis is to implement, evaluate and test a novel indoor positioning system for drone cinematography. The technology is based on infrared markers and uses visual odometry for localization. In a first step, the student should re-implement and test the system using offline data. In a second step, the estimation code should be implemented onboard on a Parrot Bebop 2 drone.
Prof. Otmar Hilliges, Tobias Naegeli (naegelit@inf.ethz.ch)
Prof. Otmar Hilliges, Tobias Naegeli (naegelit@inf.ethz.ch)