Table 6.
Summary of NIRS applications for the detection of target microbial/fungus contamination in agricultural products.
| Agricultural product | Spectral range | Software package | Number of samples | Accuracy | Findings | References |
|---|---|---|---|---|---|---|
| Tomato | 550–1,100 nm | Unscrambler v. 10.3, SpectraSuite | 45 | 74–90% | F. oxysporum f. sp. lycopersici, R. solani, B. acillusatrophaeus, and P. aeruginosa infections were detected. | (83) |
| Strawberry | 400–1,000 nm | Unscrambler X v. 10.1, SpectralDAQ v. 2.1 for STATISTICA 10 | 2,700 | 97% | B. cinerea and Collatotrichum acutatum infections were detected. The BNN model exhibited the highest predictive accuracy. | (81) |
| Kiwi | 833–2,500 nm | OPUS v. 5.5, MATLAB 2012a, Libsvm v. 3.20 | 352 | R2 = 0.961–0.999 | Z. rouxi, Hanseniaspora uvarum, and C. tropicalis infections were detected. The SVM model was on par with the plate counting method. | (6) |
| Lettuce | 350–1,100 nm | SpectraWiz, Unscrambler X10.3 | 200 | 87.1–89.39% | E. coli ATCC infection was detected. SIMCA and SVM outperformed HCA, PCA. E. coli content varied with the chemical compositions, creating non-linear relationships. | (36) |
| Cabbage | 700–1,100 nm | CA Maker, Unscrambler | 20 g | R = 0.47–0.91, SECV = 0.45–1.17 | E. coli, S. typhimurium infections were detected. Shredded leaves were more suitable for detection. Not a directly non-destructive approach. | (82) |
BPNN, Backpropagation neural network; SVM, Support vector machine; SIMCA, Soft independent modeling by class analogy; HCA, Hierarchical cluster analysis; PCA, Principal component analysis.