Volume and surface area of Holstein dairy cows calculated from complete 3D shapes acquired using a high-precision scanning system: Interest for body weight estimation

Abstract : Three-dimensional (3D) imaging is a solution for monitoring morphology and growth of dairy cows, but it can also estimate indicators such as body volume, surface area and body weight. A 3D full-body scanning device was used to scan 64 lactating Holstein cows from March-June 2018. The cows were individually and automatically weighed at a static weighing station (mean ± standard deviation = 673 ± 65 kg). These measured weights were compared to those predicted from regression models based on volume, area or morphological traits determined from 177 3D images. Since some images were truncated due to cow movement or technical problems, we developed additional regression models to reconstruct total volume or area. The accuracy of volume and area measurements was first tested on an inert cylindrical form (coefficients of variation (CVs) < 0.72%). The CVs for repeatability and reproducibility of the method of calculating volume and area from truncated images were 0.17% and 3.12%, respectively. Cow volume and area ranged from 0.61 to 0.96 m3 and 5.80 to 8.32 m2 respectively. Five regression models were developed to estimate cow body weight. Their coefficients of determination ranged from 0.82 to 0.93 with prediction errors of ca. 3% (20 kg) and 4% (29 kg) as a function of volume and area, respectively. The device and the method, evaluated and validated in this study, offer the possibility to use new indicators such as body volume and area in precision livestock farming.