Autoria: FERREIRA, I. J. S.; COSTA, D. dos S.; ROLIM, L. A.; FREITAS, S. T. de; SOUZA, N. A. C. de; TERUEL, B.

Resumo: This study aimed to investigate the use of portable NIR spectroscopy with data mining techniques for pesticide quantification in cherry tomatoes and strawberries. For each product, reflectance spectra of 240 samples, composed of three fruits and treated with different concentrations of azoxystrobin, chlorothalonil, chlorpyrifos, difenoconazole, lambda-cyhalothrin, or tetraconazole, were obtained in the wavelength range of 900 - 1700 nm, using the DLP NIRscan and FieldSpec 3 spectrometers. Reference analyses were performed using liquid chromatography. Mathematical pre-processing techniques as well as variable selection were applied to the spectral data. The regression models were developed using Partial Least Squares Regression (PLSR), Orthogonal Projection for Latent Structures (OPLS), Random Forest (RF) and Support Vector Machine (SVM) techniques. The OPLS models with selection of RFE or SFM variables were able to quantify pesticides with R2p from 0.80 to 0.96, RMSEP from 0.01 to 0.03, RPDP from 2.24 to 4.76, and R2p from 0.73 to 0.80, RMSEP from 0.06 to 0.12, RPDP from 1.93 to 2.27 in samples of cherry tomatoes and strawberries, respectively. These results show that portable NIR spectroscopy, combined with data mining techniques, holds promise for monitoring pesticide residues in cherry tomatoes and strawberries.

Ano de publicação: 2025

Tipo de publicação: Artigo de periódico

Unidade: Embrapa Semiárido

Palavras-chave: Aprendizado de Máquina, Chemical residues, Cherry tomatoes, Monitoramento de pesticidas, Morango, NIR portátil, Near infrared radiation, Pesticida, Resíduo Quimico, Strawberries, Tomate, Tomate cereja, Tomatoes