Obstacle avoidance with optic flow

Craeye, Cian Alexander (2019-04)

Thesis (MEng)--Stellenbosch University, 2019.

Thesis

ENGLISH ABSTRACT: In order to determine the potential of using optical flow as an instrument for quadcopter navigation, a real-time depth estimation system was created. The system was programmed in Python and made use of: FAST algorithm to detect key points, Lucas-Kanade pyramid algorithm to calculate the optical flow, different geometrical relationships to calculate depth from translation and Euler angles to negate the effects of rotation. A simulated testing environment created in the Unity game engine was used to successfully test the overall performance, accuracy and robustness of the system. The positive results from the system and component tests proved that there is potential in using optical flow in quadcopter navigation.

AFRIKAANSE OPSOMMING: Ten einde die potensiaal van optiese vloei as 'n instrument vir hommeltuig navigasie te bepaal, is 'n intydse diepte skattingstelsel geskep. Die stelsel is in Python geprogrammeer en maak gebruik van: FAST-algoritme om sleutelpunte te bepaal, Lucas-Kanade piramide algoritme om optiese vloei op te spoor, verskillende geometriese verhoudings om diepte van beweging te bereken en Euler-hoeke om die effekte van rotasie te negeer. 'n Gesimuleerde toetsomgewing wat in die Unity enjin geskep is, is gebruik om die algehele prestasie, akkuraatheid en robuustheid van die stelsel suksesvol te toets. Die positiewe resultate van die stelsel- en komponenttoetse het getoon dat daar potensiaal is in die gebruik van optiese vloei in hommeltuig navigasie.

Please refer to this item in SUNScholar by using the following persistent URL: http://hdl.handle.net/10019.1/106146
This item appears in the following collections: