A low-cost stereovision system to estimate size and weight of live sheep

A low-cost dual web-camera high-resolution system was developed to obtain three-dimensional position of homologous points in a scene. Two different softwares have been developed in Matlab environment. The first to acquire couples of frames by the two web-cameras, the second to process the images and obtain the points position in the space and so the distances between points. Distances between points have been corrected using a Partial Least Square regression. The efficiency of the system has been evaluated on a calibration setup experiment and then on 27 live Alpagota sheep. In the calibration phase, at a distance lower than 10 m, mean size errors were always lower than 2%. Measures on the Alpagota sheeps showed lower errors for withers height and chest depth (around 3.5%) and higher for body length (around 5.0%). These higher errors, with respect to the ones in the calibration phase, are partially due to the precision of the manual measurements on the animals. Sheep weight estimation using a PLS model on log transformed biometries showed SEP values on the data manually measured and on the stereovision extracted equal to 3.6 kg and 4.4 kg respectively. Being so inexpensive, ready-to-use and easily transportable this system could find application in other biosystems fields such as forestry, agriculture and environment. (C) 2014 Elsevier B.V. All rights reserved.