Screening of compound feeds using NIR hyperspectral data

Recent developments in spectroscopy have led to the use of spectroscopic imaging instruments for the control and monitoring of food and feed products. This kind of instruments offers the possibility of collecting thousands of spectra of particles being the result of the grinding of compound feedstuffs. The major advantages are that the recognition of feed ingredients is independent on the expertise of the analyst and that it is possible to automate all procedures and to analyse more samples per unit of time than classical microscopy or NIR microscopy. The objective of this study is the development of a new method for a rapid, precise and reliable screening of compound feeds. For that, a classification tree was built by sorting the particles in a dichotomist way where each node constitutes a discriminating step. These steps are completed by discriminant equations created from the hyperspectral databases obtained with a near infrared (NIR) camera for each class of raw materials. Discriminant equations were constructed using Support Vector Machines (SVM). For a new sample the aim is to determine its composition by using the classification tree. As general conclusion, hyperspectral data in combination with SVM as classification technique is a promised methodology for the determination of open formulations.