Poultry trials involve manually sifting through countless samples. But what if this cumbersome operation could be automated using hyperspectral imaging?
Measuring how much feed livestock ingest is a vital part of animal husbandry. The measurements are used to calculate the feed conversion ratio. However, because poultry have unique feeding habits – scratching and digging in the ground in search of feed – they are particularly difficult to study. Even when kept in cages with adequate feeding troughs, they generate feed waste, which mixes together with sawdust, droppings and other debris. As such, it is only possible to measure their feed intake accurately by identifying feed waste and separating it from other particles.
One way to do this is to sift through the samples by hand. Yet this is a laborious and time-consuming operation. Fortunately, near-infrared (NIR) hyperspectral imaging technologies mean that this process can now be automated. The samples are carefully laid out on a tray then analysed using an NIR hyperspectral imaging camera. The resulting images contain pixels, each of which corresponds to a full infrared spectrum. The spectra are then processed using chemometrics tools (via linear regression) to determine which pixels are feed waste and which are not, so that the total amount of feed waste in the sample can be measured. As well as being a real time-saver, this method also produces more reliable results than manual sorting. It could also be applied to other sample types, where one or more fractions need to be quantified