PhD Forum: GPU-based Visual Odometry for Autonomous Vehicle Applications

J. A. de Jesús Osuna-Coutiño (1) , Abiel Aguilar-González (2, 1) , Miguel Arias-Estrada (1)

1 Coordinación de Ciencias Computacionales [INAOE]
2 IP - Institut Pascal

One popular task under several computer vision applications is camera pose estimation under video sequences. In previous work, several camera pose estimations approaches have been developed and several algorithms have been proposed. Unfortunately, most previous formulations iterative behavior and depend on nonlinear optimizations applied over some geometric constraint and this limits the performance under embedded applications. For real-time embedded applications , another approach, more efficient in terms of computational size and processing speed could be reached via hardware-based solutions, for example GPU-based solutions. In this work, we present a GPU-based solution for the camera pose problem. As early formulation we focused our algorithm for an autonomous vehicle application. Preliminary results are encouraging and show the viability of the proposed approach.

Mots clés: Camera pose; embedded systems; GPU

Domaines: Vision par ordinateur et reconnaissance de formes [cs.CV]; Systèmes embarqués; Calcul parallèle, distribué et partagé [cs.DC]

J. A. de Jesús Osuna-Coutiño, Abiel Aguilar-González, Miguel Arias-Estrada. PhD Forum: GPU-based Visual Odometry for Autonomous Vehicle Applications. Proceedings of the 11th International Conference on Distributed Smart Cameras, Sep 2017, Stanford, United States. pp.2010-211, ⟨10.1145/3131885.3135083⟩. ⟨hal-01658813⟩