|Title||Growth measurements of dairy calves using a 3-D camera|
|Author(s)||Song, X.; Schutte, J.J.W.; Kamp, Arjen van der; Tol, van der, P.P.J.; Halsema, F.E.D. van; Groot Koerkamp, P.W.G.|
|Event||AgEng 2014, Zurich, 2014-07-06/2014-07-10|
Farm Technology Group
LR - Veehouderij en omgeving
|Publication type||Abstract in scientific journal or proceedings|
|Abstract||The health of dairy calves represents the herd’s future production capacity. Frequently monitoring calves’ growth can minimize losses caused by disease, death and infertility. Also, it provides the basis for a higher standard of animal welfare. A widely used variable for this monitoring is the live Body Weight (BW). However, BW being the major criterion, has its limitations. Body dimensionssuch as the height at hips, which show the skeletal development are also required as inputs for body development estimation. Inpractice, individual BW and size (e.g. chest girth) are often measured manually by farmers. This labour intensive measurement often introduces a high level distress to young cows. The objective of this research was to develop a 3D vision system to estimate a dairy calf’s body weight and skeletal dimensions automatically in order to monitor the individual growth.
Sixty-eight Holstein Friesian calves (age between one and twelve weeks) were selected at seven farms in The Netherlands in 2013. All animals were kept in group-housing systems and could freely visit an automatic calf feeding machine (CALMTM, Lely Industries N.V., Maassluis, The Netherlands). A 3D Time of Flight (ToF) camera was placed above the feeding machine horizontally (1.6 meters high). Images combined with animal identifications were recorded for six weeks. Calf’s body weight and height at hips were manually measured every week to serve as references. Locations of hipbones, tail head, body volume, rump surface and average body height were determined from the 3D body surface image. 21 image variables were created as inputs of a forward stepwise selection procedure. Based on the output from this procedure, the four most relevant variables were selected to estimate calf BW by using a multiple linear regression model. Data from forty-nine calves were selected for training the model, the other nineteen were used for validation.
As the result, the Root Mean Square Error (RMSE) of the weight estimation model was 4.26 kg with corresponding Standard Deviation (SD) of 5.30 kg (measured BWs ranged from 41 kg to 132.1 kg). The height measurements at hips had RMSEs of 2.78 cm (measured heights at hips ranged from 78 cm to 120 cm). Moreover, in weight estimations and distance measurements, there was no correlation between residuals of the prediction and references. In conclusion, it is feasible to apply the 3D vision technology to measure and monitor the calf growth automatically. These growth variables can offer not only animal health indications, but also information for future breeding selections.