Application of laser light backscattering for qualitative and quantitative assessment of dilution of clear and cloudy apple juices

This study investigated the potential of laser light backscattering imaging (LLBI) for detecting water addition in apple juices. Commercial 100 % clear and cloudy apple juices were diluted at various levels (5–50 % v/v). Backscattering images were acquired by a laser vision system equipped with a 12-bit camera and laser diodes emitting at six wavelengths in the range of 532–1064 nm. Multispectral data was extracted by signal approximation with Cauchy distribution function (M1) and first-order descriptive parameters (M2). Support vector machine (SVM) was used and the hyperparameters were optimized to maximize model performance. Coefficients of M1 achieved better classification accuracy and prediction of dilution level than those of M2. The classification accuracy increased with reduced number of output classes for both clear and cloudy juice. The binary classification of non-diluted (original juice) and diluted samples obtained the highest performance with accuracy above 87.80 %. The radial kernel utilizing M1 yielded the highest accuracy (60.00–95.00 %) for clear juice, while the polynomial kernel using M1 obtained the highest accuracy (67.50–97.56 %) for cloudy juice. Prediction of adulteration level showed the best performance with radial and polynomial kernel on clear and cloudy juice, respectively. Validation achieved R 2 = 0.615 for clear and R 2 = 0.930 for cloudy juice. The results show that the proposed technique can detect adulteration and predict dilution level of apple juices. • Clear and cloudy apple juice samples were diluted with water in 5–50 % range. • Support Vector Machine was used to classify groups as well as predict dilution level. • Cauchy function coefficients obtained better performance than descriptive parameters. • Proposed method could detect 5 % adulteration of cloudy and 10 % of clear juice.