| Hanz Cuevas Velasquez, Henrique Ferrolho |
| Thu 14 Nov 2019, 12:45 - 14:00 |
| IF, G.03 |
If you have a question about this talk, please contact: Jodie Cameron (jcamero9)
Hanz Cuevas Velasquez
Title:Segmentation and 3D reconstruction of rose plants from stereoscopic images
Abstract: The proposed method is responsible for performing the segmentation of the branches and recovering its morphology in 3D. The obtained reconstruction allows the robot to make decisions to select the candidate branches to be pruned. This method first obtains a stereo pair and calculates the disparity image using block-matching and the segmentation of the branches using a Fully Convolutional Neuronal Network modified to return a map with the probability at the pixel level of the presence of a branch. A post-processing step is carried out to combine the segmentation and the disparity in order to improve these results. Then, the skeleton of the plant and the branching structure are calculated, and finally, the 3D reconstruction is obtained. The proposed approach is evaluated with five different datasets, including interior and exterior scenes, with variable background and lighting. The different steps of the proposed pipeline are evaluated and compared with other state-of-the-art methods, showing that the accuracy of the segmentation improves other methods for this type of task, and also that the skeletonization and the reconstruction processes obtain robust results.
Henrique Ferrolho
Title:Dynamic Trajectory Optimization with Direct Transcription
Abstract: I will discuss my recent PhD research work concerning multi-contact Trajectory Optimization (TO) for high-DoF legged robots taking into account the full dynamical model of the system. Starting, I will explain how the continuous Optimal Control (OC) problem can be transcribed into a discrete approximation using Direct Transcription; here, the focus will be on a quadruped robot mounted with a robot manipulator in order to execute manipulation tasks. Afterwords, I will provide some hints on techniques for maximizing the whole-body motion robustness to external disturbances. Finally, I will address some of the remaining challenges and open-problems in this area.