Page Banner

United States Department of Agriculture

Agricultural Research Service

Research Project: Quality Based Inspection and Sorting of Specialty Crops Using Imaging and Physical Methods

Location: Healthy Processed Foods Research

Title: Detection of flaws in hazelnuts using VIS/NIR spectroscopy

Authors
item Moscetti, Roberto -
item Haff, Ronald
item Sayes, Wouter -
item Monarca, Danilo -
item Cecchini, Massimo -
item Massintini, Riccardo -

Submitted to: Journal of Food Engineering
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: March 29, 2013
Publication Date: April 8, 2013
Citation: Moscetti, R., Haff, R.P., Sayes, W., Monarca, D., Cecchini, M., Massintini, R. 2013. Detection of flaws in hazelnuts using VIS/NIR spectroscopy. Journal of Food Engineering. 118:1-7.

Interpretive Summary: The feasibility of visible light / near Infrared spectroscopy for detection of flaws in hazelnut kernels was demonstrated. Feature datasets comprising raw absorbance values, raw absorbance Ratios (Abs['1] : Abs['2]) and Differences (Abs['1] – Abs['2]) for all possible pairs of wavelengths from 306.5 nm to 1710.9 nm were extracted from the spectra for use in an iterative linear discriminant analysis routine that computed the optimal set of three said features for classification. For each dataset, several spectral pretreatments were tested. Each group of features selected was subjected to principle component analysis, receiver operating characteristics (ROC) analysis, and evaluation of performance through the area under ROC curve. The best result (5.4% false negative, 5.0% false positive, 5.2% total error) was obtained using a second Savitzky-Golay derivative on the dataset of raw absorbance differences. The optimal features were Abs[564 nm] – Abs[600 nm], Abs[1223 nm] – Abs[1338 nm] and Abs[1283 nm] – Abs[1338 nm]. The results indicate the feasibility of an economical, rapid, online detection system.

Technical Abstract: The feasibility of VIS/NIR spectroscopy for detection of flaws in hazelnut kernels was demonstrated. Feature datasets comprising raw absorbance values, raw absorbance Ratios (Abs['1] : Abs['2]) and Differences (Abs['1] – Abs['2]) for all possible pairs of wavelengths from 306.5 nm to 1710.9 nm were extracted from the spectra for use in an iterative LDA routine that computed the optimal set of three said features for classification. For each dataset, several spectral pretreatments were tested. Each group of features selected was subjected to PCA, receiver operating characteristics (ROC) analysis, and evaluation of performance through the area under ROC curve. The best result (5.4% false negative, 5.0% false positive, 5.2% total error) was obtained using a second Savitzky-Golay derivative on the dataset of raw absorbance differences. The optimal features were Abs[564 nm] – Abs[600 nm], Abs[1223 nm] – Abs[1338 nm] and Abs[1283 nm] – Abs[1338 nm]. The results indicate the feasibility of an economical, rapid, online detection system.

Last Modified: 4/17/2014
Footer Content Back to Top of Page