Object detection at different resolution in archaeological side-scan sonar images

Abstract

Side-scan sonar is now considered 'the instrument-of-choice' for underwater archaeological surveys. However, much work is required to understand the factors that may affect the surveyed data, including the effect of sonar resolution on the detection of objects on the seafloor and the potential confusion between the presence of recent objects on the seafloor and that of actual 'archaeological' material. To aid in addressing these issues, this paper presents an approach using scale saliency features for object detection in archaeological side-scan sonar images. Experimental results show that the techniques introduced here are capable of determining scales and saliency features for object regions. More importantly, the work is robust with respect to image intensity, scale and contrast. These factors often change in real sonar surveys due to collecting data at different depths and using different sonar parameters.