ANN approach for estimation of cow weight depending on photogrammetric body dimensions

Computer technology and software are widely used in every multi-discipline field. Geomatics engineering can be seen as a pioneer of these disciplines especially in photogrammetry and image processing. Photogrammetry is a method where geometric parameters of objects on digitally captured images are determined and make measurements on them. Capturing the digital images and photogrammetric processing include several fully defined stages, which allows to generate three-dimension or two-dimension digital models of the body as an end product. The aim of this study is to predict Holstein cows’ live weight via artificial neural network whose body dimensions were determined with photogrammetry method. The body dimensions to be used in this study are obtained metric from analysis of cows’ images captured by synchronized three-dimension camera environment from different aspects. Wither height, hip height, body length, hip width of cows determined with photogrammetry. Artificial neural network prediction model was developed by using these body measurements. Dataset is divided into two after preprocessing as training and testing dataset. Different structured artificial neural network models are generated and the artificial neural network model which has the best performance is determined. Then with this artificial neural network model live weight of animals is estimated by using measurements obtained from images. After comparison of estimated live weights and weights obtained from scale, correlation coefficient is found (R=0.995). The statistical analysis shows that both groups are meaningful and artificial neural network can be used in live weight prediction safely.