Development of an automated algorithm for tracking and quantifying Barnacle cyprid settlement behavior

Research focused on the development of environmentally benign antifouling technologies, novel coatings, and surfaces requires new tools for quickly and accurately evaluating these materials. Barnacle cyprid settlement assays are indispensable for estimating the fouling management potential of a coating. As coating technologies evolve to include topographies, surface active chemistries, and ligands, relying on settlement assays alone will not be enough. Understanding larval interactions and behavior with the coatings in real time is important for predicting antifouling efficacy of the coating and in turn its effects on the larvae. Manually quantifying cyprid-surface interactions from a video sequence can be laborious, highly subjective, and not amenable to high-throughput workflow requirements. Here we describe development of an automated tracking algorithm for analyzing the settlement behavior of Balanus amphitrite cyprid larvae with the aid of a lateral-view microscope equipped with a video camera. The automated algorithm uses elliptical shape models for detecting the position and orientation of larvae in the field of view and a Kalman filter based approach to automatically track the motion of the larvae over many hours of video. Behavioral heuristics are used to analyze and qualify the interaction between larvae and the surfaces. The algorithm, in an unoptimized MATLAB implementation, processes images at 11 frames per second at low median errors of 2.3 pixels and 8° for position and orientation of the cyprid respectively. The results support the use of computer vision techniques in inference and analysis of marine invertebrate larval settlement behavior.

► Complexity of current antifouling coatings needs a real-time barnacle cyprid behavioral assay.
► Used novel Lateral-view microscopy for understanding cyprid-surface interaction.
► Developed an automated cyprid tracking algorithm to track cyprids without user intervention.
► Image processing, Kalman filtering, behavioral heuristics make the algorithm flexible and robust.
► Possibility of extending proposed methods to a variety of assays.