Prediction of potato's sensory properties by Near Infrared Spectroscopy


  • Van de Laer, G. , Dardenne, P. , Agneessens, R. & Rolot, J. (1999). Prediction of potato's sensory properties by Near Infrared Spectroscopy. Poster in: 9th International Conference on Near Infrared Spectroscopy (ICNIRS), Verona - Italy, 13-18/06/1999.
Type Poster
Year 1999
Title Prediction of potato's sensory properties by Near Infrared Spectroscopy
Event name 9th International Conference on Near Infrared Spectroscopy (ICNIRS)
Event location Verona - Italy
Label 402
Recnumber 518
Event date 13-18/06/1999
Type of poster conférence
Abstract We are now facing a segmentation logic of the potato market, which differentiates the qualities according to the type of use. In these circumstances quality has become a strategical variable for producers, packers or transformers in order to get new market shares. Then the quality level of a potato share must be identified quickly. At present the culinary and technological potato quality is being determined by sensorial analyses which can be subjective and time consuming. This survey intends to show that NIRS is able to predict the culinary behaviours of potato samples. The sample set is made up of 20 potato varieties harvested in 1997 and 1998. It contains a total number of 175 samples. Ten tubers for each sample are peeled and steamed for one hour. After that, a panel of 8 trained judges quotes 5 descriptors on 11-point interval scales: disintegration, flesh colour, moisture, flouriness, and granulation. The French fries browning after frying has also been tested. For the NIR measurements, twenty tubers from each sample are washed, peeled and minced in a grinder (Hobart 8181 D) for 30 seconds. The spectra are collected by measuring the diffuse reflection of the ground sample in the VIS-IR area between 400 and 2500 nm. The spectrometer used is a NIRSystems 6500 (Perstorp Analytical Co.). Sixteen spectra are taken at different places of the measured mobile cell and 3 cells are analysed for each sample. The results of the sensorial analyses are used as referential values. Modified PLS (Partial Least Squares) is used to make the calibration on the average spectra of the three repetitions. We can see that the SECV values are close to those of the SEC. This attest the robustness of the prediction equations. As regards the colour of the tuber flesh or the frying index, a good correlation between sensorial values and NIRS predictions has been noticed. Predicting texture descriptors is less precise, but the SECV is close to the SEL, so that errors of predictions are of the same order as standard errors of the descriptors. Therefore, to improve prediction performances of NIRS models, we have to improve the sensory analysis repeatability.
Author address Dardenne Pierre, Quality Department of Agro-food Products, Walloon Agricultural Research Centre (CRA-W), Chaussée de Namur, 24, B-5030 Gembloux, dardenne@cra.wallonie.be
Fichier 3200840223402-dardenne-1999.jpg
Caption 402-dardenne-1999.ppt
Authors Van de Laer, G. , Dardenne, P. , Agneessens, R. & Rolot, J.