Bayesian Linear Regression and Natural Logarithmic Correction for Digital Image-Based Extraction of Linear and Tridimensional Zoometrics in Dromedary Camels

Abstract

This study evaluates a method to accurately, repeatably, and reliably extract camel zoo-metric data (linear and tridimensional) from 2D digital images. Thirty zoometric measures, including linear and tridimensional (perimeters and girths) variables, were collected on-field with a non-elastic measuring tape. A scaled reference was used to extract measurement from images. For girths and perimeters, semimajor and semiminor axes were mathematically estimated with the function of the perimeter of an ellipse. On-field measurements’ direct translation was determined when Cronbach’s alpha (Cα) > 0.600 was met (first round). If not, Bayesian regression corrections were applied using live body weight and the particular digital zoometric measurement as regressors (except for foot perimeter) (second round). Last, if a certain zoometric trait still did not meet such a criterion, its natural logarithm was added (third round). Acceptable method translation consistency was reached for all the measurements after three correction rounds (Cα = 0.654 to 0.997, p < 0.0001). Afterwards, Bayesian regression corrected equations were issued. This research helps to evaluate individual conformation in a reliable contactless manner through the extraction of linear and tridimensional measures from images in dromedary camels. This is the first study to develop and correct the routinely ignored evaluation of tridimensional zoometrics from digital images in animals.