This paper describes the use of particle filters for visual tracking of underwater elongated structures such as cables or pipes. Models for probabilistic tracking are obtained directly from real underwater image sequences. Extensive experiments on realistic underwater videos of power cables installed almost 30 years go (i.e. they cannot be easily discriminated from the background) show the robustness and performance of this approach. Along the paper, a model for cables as they appear in undersea images is developed within this probabilistic framework. Furthermore, real images from inspection runs captured by remotely operated vehicles (ROVs) are used to find out an appropriate movement model which describes the changes of the cable parameters over time. Finally, an observation model which detects edges by a derivative of a Gaussian filter is also defined. All these models are combined in a particle filter which sequentially estimates the likelihood of the cable pose. For every frame in the video sequence, the previously computed probability density function of the cable parameters is used to predict - via application of the movement model - the cable pose in the next frame. Then, the probability density function is updated by means of the observation model. The most likely cable pose is finally determined from the resulting density. The paper is organized as follows: section 2 gives a brief introduction to particle filters; models for probabilistic cable tracking are developed in section 3; section 5 outlines our experiments with real underwater image sequences
Using particle filters for autonomous underwater cable tracking
Einsatz von Partikelfiltern für the autonome Verfolgung von Unterwasserkabeln
2008
6 Seiten, 9 Bilder, 1 Tabelle, 6 Quellen
Conference paper
English
Underwater Multi-Target Tracking with Particle Filters
IEEE | 2018
|Cable Tracking of Autonomous Underwater Vehicle "Aqua Explorer 1000"
British Library Conference Proceedings | 1994
|Subsea cable tracking in an uncertain environment using particle filters
Taylor & Francis Verlag | 2015
|Underwater mobile target tracking with particle filter using an autonomous vehicle
BASE | 2017
|