Metodi per l'elaborazione, il riconoscimento e l'integrazione automatica di mappe ottiche ed acustiche subacquee = Methods for automatic processing, recognition and integration of underwater optical and acoustic maps

Moroni D, Pascali M A, Reggiannini M, Salvetti O

Image understanding  Texture analysis  Structure from motion  I.2.10 Vision and Scene Understanding  I.4.8 Scene Analysis 

Data acquired from optical and acoustic sensors placed on autonomous underwater vehicles are processed by means of automated Computer Vision procedures. Photometric distortions in optical images are rectified by calibration of the optical system, whereas Slant/Ground Range algorithms applied to acoustic maps correct the geometrical distortions typically arising in acoustic detection. Environmental noise affecting both data kinds are removed implementing appropriate filters. Furthermore, the signal is processed by maps registration and geo-referencing algorithms, features and pattern recognition, texture recognition and classification algorithms. Three-dimensional seabed reconstruction from optical images (Structure From Motion) and acoustic maps (Shape From Shading) allows a straight comparison between the two 3D models describing the same scene. A procedure returns the fusion of the two data types.

Publisher: CNR