Publication: Interfaz y librería para visión artificial, navegación y seguimiento en robótica
Loading...
Identifiers
Publication date
2012-05
Defense date
2012-07-03
Authors
Tutors
Journal Title
Journal ISSN
Volume Title
Publisher
Impact
Export
Abstract
En este proyecto se han desarrollado un conjunto de algoritmos de visión artifical, que permiten a los robots el seguimiento de objetos, la detección de movimiento y la navegación por entornos no conocidos. En cuanto a los algoritmos de seguimiento, se han utilizado técnicas ya conocidas como SURF, meanshift o template matching y un m etodo propio: Colormax. La detección de movimiento se realiza a través de optical flow. Para la navegación se han creado dos métodos nuevos para arquitecturas reactivas. NaviOtsu, que se basa en la umbralización por el método Otsu para evitar obstáculos, y NaviColor, que utiliza la distribución de colores del entorno para elegir el camino más adecuado. La implementación de los algoritmos está basada en la librería OpenCV y estos se han integrado dentro de una interfaz multiplataforma que simplifica su uso y permite que sean utilizados por cualquier robot. Se incluyen también estadísticas y comparativas entre los algoritmos desarrollados para comprender las ventajas e inconvenientes que ofrece cada uno. Por último, se ha realizado una demostración del software en un pequeño robot humanoide y se han comprobado los resultados de utilizar los algoritmos en una plataforma real. _____________________________________________________________________________________________________________________
In this project there have been developed a set of vision algorithms that allow object tracking, motion detection and navigation. The algorithms of tracking have been designed using popular techniques like SURF, meanshift or template matching and one new method: Colormax. Motion detection is achieved by using optical flow. There have been designed two new methods for reactive navigation architectures. NaviOtsu, which is based on Otsu thresholding to avoid obstacles and NaviColor, that uses environment colors distribution to choose the best appropriate path. Algorithms implementation are based on the OpenCV library and they have been integrated into a multi-platform interface, which makes more friendly and useful their use in robots. It also includes statistics and comparisons between algorithms and their advantages and disadvantages. Finally, a software demostration has been accomplished by using a small humanoid robot and have proven the results of using the algorithms in real environtments.
In this project there have been developed a set of vision algorithms that allow object tracking, motion detection and navigation. The algorithms of tracking have been designed using popular techniques like SURF, meanshift or template matching and one new method: Colormax. Motion detection is achieved by using optical flow. There have been designed two new methods for reactive navigation architectures. NaviOtsu, which is based on Otsu thresholding to avoid obstacles and NaviColor, that uses environment colors distribution to choose the best appropriate path. Algorithms implementation are based on the OpenCV library and they have been integrated into a multi-platform interface, which makes more friendly and useful their use in robots. It also includes statistics and comparisons between algorithms and their advantages and disadvantages. Finally, a software demostration has been accomplished by using a small humanoid robot and have proven the results of using the algorithms in real environtments.
Description
Keywords
Visión artificial, Robótica, Proceso de imágenes, Algoritmos, Interfaces de usuario, OpenCV