Exploring 3-D Reconstruction Techniques: A Benchmarking Tool for Underwater Robotics

Abstract

Abstract-Nowadays, when the results of research in the field of robotics are presented to the scientific community the same question is asked repeatedly: are the results really reproducible? Regarding benchmarking issues, some technological areas, where complex mechatronic devices such as robots have a central role are, in general, very far from other research areas like physics or chemistry, to name but a few, where reproducibility is always mandatory. Leaving aside mechatronic complexities, the comparison between two different algorithms in the same conditions is influenced by the experimental validation scenario. In underwater environments, the difficulties for benchmarking characterization increase substantially. This is especially true when the testbed is the sea where uncertainty is really high. It is the aim of this work to present a software tool which enables a comparison between two different algorithms to be made when these algorithms are being used to solve the same problem in water tank conditions. This is a preliminary stage before the final validation on the seabed. The evaluated algorithms fall into the 3D image reconstruction context, as a prior step to their autonomous manipulation. Performance results are presented for both simulation and real water tank conditions.

Integración de una cámara pan-tilt subacuática para detección de objetos y guiado de un ROV submarino

“…Es bien conocido que en elámbito marino, la localización mediante dispositivos tipo sonar (pasivo o activo), es una de las principales herramientas para detección y evasión de colisión de los diferentes obstáculos que se puedan encontrar sumergidos bajo la superficie del agua, que si bien, proporcionan una reconstrucción tridimensional del mapa marino [3], no están exentos de un altoíndice de ruido, a la vez que puede afectar al comportamiento de la fauna que habita en el entorno del radio de acción, como se muestra en [5], ya sea emitiendo tanto ondas electromagnéticas como acústicas. A parte de estas ventajas e inconvenientes, se encuentra la necesidad que nos atañe en este documento, la detección de objetos de interés (2D o 3D), tanto de forma autónoma como supervisada [10], donde las cámaras son sensores especialmente adecuados.…”

UWSim, un simulador submarino conectado a la nube como herramienta educacional

“…En función de unas métricas configurables dependiendo del objetivo de la intervención, evalúa el resultado de la misma. Además, en el caso de calibrar correctamente el entorno virtual del simulador con el real, es posible evaluar algoritmos en entornos reales como se describe en Perez et al (2015).…”

Monocular Robust Depth Estimation Vision System for Robotic Tasks Interventions in Metallic Targets

“…Sensor fusion systems can make use of cameras with depth information like Kinect [16] and RealSense [17]. Additionally, some related works make use of stereo cameras to track a target, in which both cameras are placed at a predetermined distance and rotation, calculating a whole 3D reconstruction of the scene [18,19,20] using Epipolar Geometry [21]. A single-camera system can also be deployed [22] simulating a stereo system, either using markers [23], or previously establishing the separation parameters among two images [24], knowing the relationship between the key-points of both images, necessary to build the Epipolar Geometry.…”