Skip to main content
ARS Home » Southeast Area » Athens, Georgia » U.S. National Poultry Research Center » Quality and Safety Assessment Research Unit » Research » Publications at this Location » Publication #315937
Title: Classification of specialty seed meals from NIR reflectance spectra

Author
item Holser, Ronald
item Kandala, Chari
item PUPALA, NAVEEN - New Mexico State University

Submitted to: Trends in Applied Spectroscopy
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 1/19/2015
Publication Date: 2/26/2015
Citation: Holser, R.A., Kandala, C., Pupala, N. 2015. Classification of specialty seed meals from NIR reflectance spectra. Trends in Applied Spectroscopy. 11:5-18.

Interpretive Summary: The food and animal feed industries rely on automated quality control systems to quickly and non-destructively evaluate natural products and other ingredients with variable compositions. The combination of infrared reflectance measurement with classification algorithms was shown to reliably identify and sort a variety of seed meals for use in animal feed formulations. These results allow formulators to confidently use alternative components and achieve the desired nutritional quality in products. This technique is expected to be widely accepted by industry.

Technical Abstract: Near infrared reflectance spectroscopy was used to identify alternative seed meals proposed for food and feed formulations. Spectra were collected from cold pressed Camelina (Camelina sativa), Coriander (Coriandrum sativum), and Pennycress (Thlaspi arvense) meals. Additional spectra were collected from Dried Distillers Grains with Solubles (DDGS) which is an inexpensive co-product obtained from dry milling maize (Zea mays). Spectra were processed by multiplicative scatter correction and first derivative transform prior to principle component analysis (PCA). The PCA score plots showed separate groups for each material and identified potential groups for classification by linear discriminant analysis (LDA), soft independent modeling of class analogy (SIMCA), and support vector machine (SVM) methods. Results showed that LDA and SVM were both successful in classifying the type of source material while SIMCA was not able to correctly identify solvent extracted materials. The ability to rapidly and non-destructively confirm the identity and quality of components at the process line will promote the use of alternative seed meals to supplement commodity meals such as maize and soybean.