3D Pose Estimation and Tracking of an Electricity Pylon
Emmanuel Y. Ali, Fred Nicolls · 2020 International SAUPEC/RobMech/PRASA Conference · 2020
We demonstrate an algorithm for estimating the 6 degree-of-freedom pose of a textureless object such as a power pylon. The method is designed to be part of the inspection process for a transmission line. The approach involves three steps, namely predicting the vertices of the pylon with a neural network, establishing the correspondence between vertices with hashing techniques, and incrementally tracking the movement of the camera. We built a pylon model for the experiments and used it to generate our dataset.