Computer-Assisted Analysis of Near-Bottom Photos for Benthic Habitat Studies

摘要

This paper reports on a methodology developed for the analysis of near-bottom photographs collected for fisheries habitat studies. These tools provide a framework for conducting minimally invasive in-situ investigations of benthic organism abundance, diversity, and distribution using high-resolution optical datasets integrated with high precision navigational data. Utilizing these techniques with near-bottom photos collected with a precision navigated survey platform greatly increases the efficiency of image analysis and provides new insight about the relationships between benthic organisms and the habitats in which they are found. Basic requirements for the analysis of near-bottom seafloor images include camera calibration and quantification of the height of the lens above the seafloor throughout the survey. Corrections are required to compensate for image distortion due to lighting limitations and the variable micro-topography of the seafloor. These parameters can be constrained by utilizing precisely navigated survey platforms such as Autonomous Underwater Vehicles (AUVs) or Remote Operated Vehicles (ROVs). The methodology we present was developed with data collected by the SeaBED AUV off the coast of Washington, Oregon and California [1]. A digital database containing benthic organism identifications, measurements, and locations was generated for each image using a Graphical User Interface (GUI) created in Matlab(tm) [1,2]. This methodology has demonstrated a significant increase in the efficiency of image analysis for benthic habitat studies, and provides the opportunity to assess small scale spatial distribution of organisms in their natural habitats. Collecting overlapping images permits the creation of photomosaics [3] and the quantification of organism abundance per unit area of the seafloor.