Body volume and mass estimation of southern elephant seals using 3D range scanning and neural network models

Abstract

Direct measures of body mass of marine mammals are logistically complicated to obtain even for pinnipeds. An alternative method for mass estimation based on 3D imaging technology and automated processing algorithms, was tested in southern elephant seals (Mirounga leonina). Two models of artificial neural networks (ANN)—nonlinear neural network and self‐organizing maps—were trained to compute the volume and to estimate the mass of the digital models, previously obtained by scanning individuals with an infrared light sensor. Body mass estimates were as accurate (mean % error = 4.4) as estimates in previous photogrammetry studies in southern elephant seals and the mass predictive ability of the trained ANN was higher (99% of the variance explained) than other predictive models using photogrammetry in pinniped studies. While this method has proven to produce accurate body mass estimates, it also overcame some of the constraints of other indirect techniques, avoiding animal disturbance caused by physical restraint or chemical immobilization, minimizing risks, and capitalizing on the time working in the field. The results of these estimations were promising, which shows that the proposed methodology can provide adequate results with lower logistic and computational requirements.

WOLO: Wilson Only Looks Once – Estimating ant body mass from reference-free images using deep convolutional neural networks

“…As visual information alone is often insufficient to accurately estimate size, it is common practise to include absolute scale information in the images [1,3,7,9], for example in form of reference objects of known size. Radar, sonar, or infrared light, can help address the same problem, because weight can then be estimated from coarse 3D object reconstructions [10][11][12]. Completely reference-free size-estimation, however, is rare [6].…”

Marine mammal morphometrics: 3D modeling and estimation validation

“…However, hosting intact bodies of marine mammals into scanners, often existing in hospitals, is challenging. Surface scanning uses the reflection of structured light (e.g., laser) to construct 3D models and this technique has been explored with several species of stranded pinnipeds (Ciobanu et al, 2013;Beltran et al, 2018;Eder et al, 2022). Attempt applications of such technique to animals at sea is complex due to stronger noise signal from water surface than reflection patterns from target animals (Czepkowshi and Słoẃko, 1996;Heritage and Hetherington, 2007;Suchocki and Katzer, 2018).…”