So far, there has been no safe and convenient method to weigh the large fierce animals, like Amur tigers. To address this problem, we built models to predict the body weight of Amur tigers based on the fact that body weight is proportional to body measurements or age. Using the method of body measurements, we extracted the body measurements from 4 different kinds of the lateral body image of tigers, that is, total lateral image, central lateral image, ellipse fitting image, and rectangle fitting image, and then we respectively used artificial neural network (ANN) and power regression model to analyze the predictive relationships between body weight and body measurements. Our results demonstrated that, among all ANN models, the model built with rectangle fitting image had the smallest mean square error. Comparatively, we screened power regression models which had the smallest Akakai information criteria (AIC). In addition, using the method of age, we fitted nonlinear regression models for the relationship between body weight and age and found that, for male tigers, logistic model had the smallest AIC. For female tigers, Gompertz model had the smallest AIC. Consequently, this study could be applied to estimate body weight of captive, or even wild, Amur tigers safely and conveniently, helping to monitor individual health and growth of the Amur tiger populations.
Keywords: Amur tiger; Panthera tigris altaica; body weight; growth curve; image analysis.
© 2021 International Society of Zoological Sciences, Institute of Zoology/Chinese Academy of Sciences and John Wiley & Sons Australia, Ltd.