Efficient artifacts filter by density-based clustering in long term 3D whale passive acoustic monitoring with five hydrophones fixed under an Autonomous Surface Vehicle

摘要

Passive underwater acoustics allows for the monitoring of the echolocation clicks of cetaceans. Static hydrophone arrays monitor from a fixed location, however, they cannot track animals over long distances. More flexibility can be achieved by mounting hydrophones on a mobile structure. In this paper, we present the design of a small non-uniform array of five hydrophones mounted directly under the Autonomous Surface Vehicle (ASV) Sphyrna (also called an Autonomous Laboratory Vehicle) built by SeaProven in France. This configuration is made challenging by the 40cm aperture of the hydrophone array, extending only two meters below the surface and above the thermocline, thus presenting various artifacts. The array, fixed under the keel of the drone, is numerically stabilized in yaw and roll using the drone's Motion Processing Unit (MPU). To increase the accuracy of the 3D tracking computed from a four hour recording of a Sperm Whale diving several kilometers away, we propose an efficient joint filtering of the clicks in the Time Delay of Arrival (TDoA) space. We show how the DBSCAN algorithm efficiently removes any outlier detection among the thousands of transients, and yields to coherent high definition 3D tracks.