Automatic Body Condition Scoring on Dairy Cows of the Swedish red breed

Sammanfattning: The objective of this MSc thesis was to investigate parameters reported to be associated with dairy cows’ body condition and its changes. A valid reference of a dairy cow’s body condition was to be suggested and investigated for suitability in the development of a 3D imaging based automatic body condition scoring model. Furthermore, it was important to accumulate data to use in training (adjustment of the algorithm mathematics) of the 3D imaging based automatic body condition scoring model. The study included 21 dairy cows of the Swedish Red breed from the herd at the Kungsängen Research Centre in Uppsala. The cows had access to an exercise pen and were fed silage and concentrates indoors according to the Swedish feeding recommendations, based on individual milk yields. Data was collected weekly from May to August 2009 and included live weight, manual body condition score, backfat thickness, 3D images, milk yield, content of fat, protein and lactose in milk and the plasma metabolites non esterified fatty acids and β-hydroxybutyrate. Data was analysed by linear correlation and regression analysis. Of all individually investigated collected parameters, backfat thickness was found to have the highest correlation with manual body condition scores and this parameter was therefore suggested and used as an alternative true reference of body condition in the training of the 3D imaging based automatic body condition scoring model. Results were promising and it was concluded that it is possible to train and calibrate the 3D imaging based automatic body condition scoring model both with manual body condition scores and with backfat thickness as reference, to predict the dairy cow’s body condition. The advantage in using backfat thickness would be that it is a more objective measure and that it gives continuous data instead of the categorical data obtained from manual body condition scoring. If sufficient sensitivity is obtained in the automatic body condition scoring model it could alert the farmer to changes in the cow’s body condition, instead of only recording the body condition after a change. Future studies should focus on developing this function in the automatic body condition scoring models since it would add significant value to dairy management systems.