Sonar Point Cloud Processing to Identify Sea Turtles by Pattern Analysis

Abundant in coastal areas, sea turtles are affected by high-intensity acoustic anthropogenic sounds. In this article, we offer a pattern-analysis-based detection approach to serve as warning system for the existence of nearby turtles. We focus on challenge overcoming low signal-to-clutter ratio (SCR) caused reverberations. Assuming that, owing SCR, target reflections within point cloud received groups, our detector searches patterns through clustering identify possible “blobs” reflections, and classify them either clutter or target. Our unsupervised is based geometrical spectral constraints over blob's member relations. turn, classification identified blobs features extracted from reflection pattern. To end, assuming that turtle stable but include diversity due distortions turtle's body, quantify stability members entropy their spectrum. test both modeled simulations, at sea, released after rehabilitation. The results show robustness highly fluctuating intensity ability detect SCR.