Acoustic Positioning Enhanced Optical Image Mosaics: A Collaborative Robot-Assisted Approach

While underwater acoustic imaging devices are im-mune to issues such as turbidity and low visibility, the quality of information that an optical camera provides is richer, if they are in close proximity to an object of interest. Visual inspection is common for large underwater structures such as pipelines, where videos and image mosaics are prepared since individual images may not represent the structure. Current works focus on enhancing individual images and utilizing several techniques to determine features across the sequence of images to stitch them into a mosaic. Given the advances in non-visual sensor-fusion for navigation, dependence of feature-based methods for mosaic generation can be avoided as challenging underwater environments may complicate feature extraction. Therefore, this study aims to generate image mosaics purely utilizing navigation information from a linear Kalman filter (KF), which fuses Inertial Measurement Unit (IMU) and SONAR data; a single-beam SONAR attached on an underwater vehicle (UWR) and forward-looking SONAR (FLS) observing the UWR. From experiments, image mosaics were successfully prepared without the use of any feature-based methods, from a distance of $1m$ from the target, with the mosaic being about $1.5 m$ in width, aided with navigational data from the KF.