Construction of a Data Integration Platform for the Passive Monitoring of the Antillean Manatee in Panama

In ecology, there are several strategies used for gathering samples in orde to understand behavioral patterns of an animal, or to quantify their population size. In the case of endangered species, techniques such as, mark and recapture, becomes impossible. Thus, researchers collect their data in passive ways mostly, bioacoustic, or visual monitoring. In this work we present elements for the construction a data integration platform geared to structure the acoustic and visual samples obtained from the passive monitoring of the Antillean Manatee, an endangered species of marine mammals that inhabit rivers and brackish waters of Panama. The article present an overview of the methods and algorithms featuring signal processing and deep learning for its analysis. Methods that include applications for the acoustic detection, classification and clustering of manatee vocalizations are presented with partial results considering the current software implementations. Additionally, an approach is described applying computer vision algorithms for the detection of manatees in images extracted from video footage shot from unmanned aerial vehicles (UAV).

I. Introduction

With the rise of technologies based on Big Data (BD) and cloud computing or Cloud Computing (CC), the development of data analysis platforms in the area of bioacoustics has been encouraged in recent years. One of the first works corresponds to the B2p2 platform developed by Thudumu et al. [1] oriented to the efficient processing of large audio files using popular Big Data libraries such as Apache Spark and Hadoop implemented in the Scala programming language. Subsequently, Brown et al. [2]proposed a scheme for scalable preprocessing for monitoring bird species and then developed the platform called AcoustiCloud with a CC-based infrastructure, integrating the preprocessing and data analysis stages [3].