Estimation of Optical Properties in Postharvest and Processing Technology

Non-destructive analysis and qualification methods are of great interest in agriculture, postharvest technology and food processing. Computer vision systems have the additional advantage that sensors do not touch the product and measurements can be performed from comfortable distance. Due to the recent developments in electronics, several commercial applications are already available for grading on the basis of visible attributes, near infrared (NIR) readings (GREEFA, The Netherlands; MAF Industries Inc., USA) and laser scattering (BEST NV, Belgium). Portable devices also exist providing optical quality assessment for flexible measurements and in vivo inspections (CP, Germany; Unitec S.p.A., Italy). Information about interaction between light and biological tissue is essential in visual evaluation of fresh horticultural produces, raw materials and food, since optical signal is significantly affected by physical stage and valuable compounds of the tissue. Hidden physical damages in cucmber fruit, caused during harvest, transport and handling, were investigated in the spectral region of 950-1650 nmusing hyperspectral imaging system (Ariana et al., 2006). Four specific wavelengths were selected for classification. The ratio of relative reflectances calculated as 988 nm to 1085 nm and the differences of this property obtained using 1346 nm and 1425 nm resulted in the highest classification rates. This study also confirmed that time plays very important role in qualification of perishable produces. Light transmittance through cucumbers in the range of 500-1000 nmwas also investigated (Ariana & Lu, 2010). Transmittance for internally defected pieces and pickles was found to be generally higher compared to normal cucumbers. The wavebands around 745, 765, 885 and 965 nmwere selected for best detection accuracy (94.7%). Advanced statistical methods, such as partial least square discriminant analysis (PLSDA) and k-nearest neighbor (KNN) may utilize the whole transmitted spectra. The highest classification rates of 97.3% and 88% were reached using PLSDA and KNN, respectively. Besides detection of damages and mechanical injury, estimation of quality parameters is also important for prediction of shelf-life and grading. Key quality attributes, mainly firmness and soluble solids content (SSC), were predicted for apple fruits based on multispectral imaging (Lu, 2004; Qing et al., 2007; 2008). Firmness and SSC were predicted for ’Red Delicious’ apples, using the ratio of backscattering profiles in the wavelength range of 680-1060 nm, with the standard error of prediction (SEP) 5.8 N and 0.78%, respectively (Lu, 2004). Four specific wavelengths (680, 880, 905, and 940 nm) were selected for firmness and three (880, 905 and 940 nm) for SSC prediction. The size of the diffusively illuminated surface area and statistical description of the acquired 34