Plane Based Visual Odometry for Structural and Low-Texture Environments Using RGB-D Sensors

This paper exploits planes instead of feature points, and proposes a visual odometry approach to estimate sensors' poses for structural and low-texture environments. We construct constraints for factor graph by utilizing the associated planes based on line features, and we derive the analytic jacobian for planes' minimal representation to speed up the optimize process. Experiments demonstrate the effectiveness and robustness of our proposed method and show its advantages over alternative methods.