Investigation of Vision-Based Underwater Object Detection with Multiple Datasets

Abstract

In this paper, we investigate the potential of vision-based object detection algorithms in underwater environments using several datasets to highlight the issues arising in different scenarios. Underwater computer vision has to cope with distortion and attenuation due to light propagation in water, and with challenging operating conditions. Scene segmentation and shape recognition in a single image must be carefully designed to achieve robust object detection and to facilitate object pose estimation. We describe a novel multi-feature object detection algorithm conceived to find human-made artefacts lying on the seabed. The proposed method searches for a target object according to a few general criteria that are robust to the underwater context, such as salient colour uniformity and sharp contours. We assess the performance of the proposed algorithm across different underwater datasets. The datasets have been obtained using stereo cameras of different quality, and diverge for the target object type and colour, acquisition depth and conditions. The effectiveness of the proposed approach has been experimentally demonstrated. Finally, object detection is discussed in connection with the simple colour-based segmentation and with the difficulty of tri-dimensional processing on noisy data.