Title:  Robust Monocular Visual Odometry For Outdoor Robotic Applications

Committee: 

Dr. Pradalier, Advisor

Dr. Vela, Co-Advisor    

Dr. AlRegib, Chair

Dr. Egerstedt

Dr. Dellaert

Abstract:

The objective of this thesis is to explore the potentials to enhance the visual odometry (VO) robustness and accuracy for outdoor robotic applications. The typical challenges of VO for natural environmental monitoring are addressed, where the robust VO algorithms are developed to improve the tracking and mapping robustness of the monocular system in the presence of large scale variance, illumination changes, and fast camera motions. Besides, the successful implementation of robust VO for robotic weed control application is presented to support our claim further. Finally, the periodic summary and plan are exhibited.