NIR goes shopping: miniaturized NIR equipment for daily tasks.

NIR spectroscopy has been a technique used mainly in laboratories and industrial plants, and still away from the lay people. New advances in electronic areas, such as in microelectroptical systems, made it possible to create an equipment for NIR analysis with highly reduced size and cost, making it portable and able to be carried around. Therefore, time has come to introduce it to the non-specialized citizen, so it can access NIR vast applications and fast analysis in order to facilitate and help in daily tasks, such as shopping. With highly reduced cost and a friendly interactive platform, it is possible to take the NIR technique to the everyday common non-specialized citizen or local market, so it could be used to help a customer to select a product of according to his own taste or even check its authenticity. The use of MicroNIR 1700, from JDSU Company, along with chemometrics can result in a powerful tool to build up in this vision. In this way, a prototype for customer assistance is proposed. The prototype incorporates chemometrics models so it can be used to pretreating and analyze all data obtained from the MicroNIR almost instantly, providing in virtually real time the information sought by the consumer.

The MicroNIR has highly reduced dimensions (45mm x 42mm), can obtain a NIR reflectance spectrum in the range 900 – 1700 nm, and was evaluated in 2 assays for proof of concept. The first assay focus on the use of NIR spectra sets to construct models to identify fabrics using PCA models. The second assay uses the MicroNIR together with data obtained from a refractometer to build PLS models to determinate the BRIX of apples, kiwis and oranges fruits, reflecting directly the sugar content in the fruit. The models will be encompassed on the prototype, so it can provide all the information the customer wants to know about the product being purchased.

For assay 1, spectrum data from 100 samples of different fabric materials (cotton, polyester, silk and mixed fabrics) were gathered and pre-treated with SNV. Them a PCA model were made, resulting in PC1 and PC2 explaining 95% of the spectral variability. The different fabrics materials clustered in very distinct groups in PC1 x PC2 graph. SIMCA models were validate with average 98 % of correct classification. Therefore, SIMCA models can be incorporated in the prototype, enabling the consumer to verify if the fabric material displayed in the labels and/or claimed by the seller is, in fact, correct. For assay 2, panels with 100 sample of each; apples, kiwis and oranges, were gathered, and their BRIX values and NIR spectrum were obtained. PLS models were constructed, for each different type of fruit, for determination of BRIX without harming the sample. All results present an RMSEP lower than 0.7 BRIX unit, and criteria to classify the fruits based on BRIX threshold values were proposed, incorporated into the prototype, and validated by correctly classifying the fruits in sweet, acid, and tasteless in accord with a sensorial panel.