University of Helsinki - A Leading European University

News of the week

Week 26/2007: Neural network model
combined with milking robot identifies lameness

As farms and herds increase in size, new machines and appliances have been introduced to make farming easier and more effective.

One such invention is a milking robot for a loose-housing system, into which the cows go to be milked 1 to 3 times a day. The idea is that the cow enters the milking stall whenever she feels like it. Sometimes she is encouraged to do so by giving her some extra feed.

The cow wears a transponder which helps the milking robot to identify her. The same kind of identification is also used in the measurement software designed by Matti Pastell, M.Sc. (Agr.).

His doctoral dissertation is entitled Automatic Lameness Detection in a Milking Robot: Instrumentation, measurement software, algorithms for data analysis and a neural network model.

Lameness is detected using four scales, which are situated under the platform of the milking robot and which measure how much weight the cow puts on each foot.

The computer program starts measuring once the cow enters the stall and registers the weight of each foot during milking. Based on these data the program calculates the ratio between the weights in per cent and the neural network determines whether the cow is healthy or not.

“If one leg delivers less weight than the others, there is reason to suspect that the cow has a sore hoof”, Pastell explains.

The measurements were carried out on 74 cows at the Suitia research farm. Data for 10,000 milkings were registered by the program. These data were then combined and the neural network was ‘taught’ 5,000 healthy and non-healthy measurements.

Later Pastell designed an error correction system for situations in which the cow stands slightly off-centre.

The result was a general model which makes it possible to determine whether a cow is healthy or not. In 96 per cent of the cases the program made the correct decision. The i nformation is fed directly to the farmer’s computer.

Lameness is a very common problem with cows. In addition to causing pain to the animal, it also means extra costs for the farmer.

“You get less milk, more medical costs and, if the worst comes to the worst, the animal may have to be put down”, says Pastell.

If the ailment is noticed at an early stage, the loss of output is less and the medical costs are normally lower too.

"Most hoof illnesses affect the hind legs, and therefore the neural network has been programmed to compare the cow's hind legs."

 

Text: Ilona Hietanen
Photo: Matti Pastell
www.helsinki.fi/digitalcommunications
Translation: Valtasana Oy

Research news archive»»