Abstract:
Currently, there are two popular methods for determining the weight of a cow: 1) Using a weighing machine which high accuracy 2) Using a measuring tape around the cow's chest. This method will be convenient, but there is quite a high degree of error. Both
methods mentioned must have direct contact with the cow. Therefore, it might be dangerous for man would be unskilled or familiar with cows. This thesis therefore presents using of artificial neural networks combined with trigonometry theory for estimating the
weight of a cow from images taken via a smartphone camera on the Android operating system. Searching the position of a cow's chest and back are using a convolutional neural network to measure the chest size and height of a cow by applying trigonometry theory.
Weight was assessed using a polynomial regression equation from the relationship between weight, chest size, and height of the cow. Data were collected from the cattle farm of Faculty of Agricultural Innovation and Technology at Nong Rawiang Educational
Center, Rajamangala University of Technology Isan, Nakhon Ratchasima Province. The experiments were carried out on large and small Brahman cows. The photography environments included both closed pens with controlled lighting and open (outdoor)
pens with natural lighting conditions. A total of 23 cows were used and 765 photographs in this experiment. The weight results of cows in open pens were accurate at 91.306% and in closed pens at 90.832%, respectively. It is an important factor of the photograph
to achieve accurate results must be a side view and all features of the cow are completely visible, such as the feet, chest, and back. However, the data for evaluating cow weight in this thesis was done on Brahman cows only. Such principles can be applied
to cows of other breeds in the future