Multiple‐angle acoustic scattering and classification of zooplankton

Abstract

The use of multiple‐angle acoustic scattering to discriminate between two ecologically important classes of zooplankton’ copepods and euphausiids, is proposed. The distorted wave Born approximation is used to model the multiple‐angle scatter from a linear array of transducers designed to be practical for field deployment. Using a large set of noisy training and test data generated from uniformly random length and three‐dimensional orientation distributions from each class of scatterers, the performance of classification algorithms in several feature spaces is evaluated. The results show a marked improvement in classification performance as additional angles, and larger angular separations, are included in the classifier. Interestingly, even in the case of uniformly random three‐dimensional scatterer orientation, low classification error (∼5%) can be obtained. These results hold promise for substantially improving the classification of fluidlike zooplankton in situ using multiple angle scatter obtained with a simple collection geometry.