The main statement that this work pretends to defend is: It is possible to track the motion of a robot with moderated accuracy using a single fisheye
camera and common sensors on top of a low-cost microprocessor in an outdoor environment. The proposed system aims to localize in real-time a ground robot in an agricultural environment. As shows in Figure 1, a fusion of common sensors with a monocular omnidirectional Visual Odometry (VO) algorithm is performed.
The entire system runs on top of a low-cost microprocessor, a Raspberry Pi 3B. An OpenCL-based optimization approach applied to the VO method is proposed recurring to the Raspberry Pi’s GPU to overcome the performance limitations of this microprocessor. The central unity of the system is the VO method. This one is publicly available on the official Libviso2 repository (https://github.com/srv/viso2) and can work standalone, giving a primary estimation of the robot motion. A sensing system,constituted by a planar laser and a gyroscope,is also proposed to support and solve the main limitations monocular VO. The first is used to calculate the motion scale due to the unavailability of depth information resultant from a monocular VO system. The last is used as a support to VO in rotations and it is fused with VO recurring to a Kalman Filter (KF). The system output is a homogeneous transformation [R|t] between consecutive image frames.
André Silva Aguiar, Armando Jorge Sousa, Filipe Neves dos Santos
Faculdade de Engenharia da Universidade do Porto
INESC TEC – Instituto de Engenharia de Sistemas e Computadores, Tecnologia e Ciência
L. FAST-FUSION: An Improved Accuracy Omnidirectional Visual Odometry System with Sensor Fusion and GPU Optimization for Embedded Low Cost Hardware. Appl. Sci.2019, 9, 5516.
Para ler o artigo completo faça o download gratuito da revista “robótica” nº119. Verifique o seu email após o download. Pode também solicitar apenas este artigo através do email: a.pereira@cie-comunicacao.pt
Outros artigos relacionados
- Artigo “Visão artificial na criação automatizada de micropeças personalizáveis” da edição 119 da revista “robótica”;
- Artigo “Omnidirectional Robot using Internet of Things (IoT)” da edição 108 da revista “robótica”;
- Artigo “Robot para Busca e Salvamento baseado na Plataforma Raspberry PI 2” da edição 102 da revista “robótica”;