Identification of tomatoes with bruise using laser-light backscattering imaging technique

Abstract

Accurately detecting bruise in tomatoes is a significant challenge in postharvest processing due to large-scale production and the demand for high-quality standards. Reduced nutritional value and product appearance due to bruise can affect market value and selling price. Therefore, this study aimed to develop a sensing technology based on laser-light backscattering imaging (LLBI) to identify the bruise of tomatoes. A total of 300 samples were collected and labeled into two main classes namely sound and bruised tomatoes. Each fruit was scanned with various laser wavelengths (450, 532, and 648 nm) at an incident angle of 20° to monitor the structural characteristics. In backscattering image, the gray-level co-occurrence matrix was implemented to extract six texture features including contrast, dissimilarity, homogeneity, energy, angular second moment, and correlation. Additionally, a support vector machine with various kernel functions namely linear, radial basis function (RBF), and polynomial was used to detect bruise. The results showed that based on numerical analysis, the LLBI technique was capable of identifying bruise in tomatoes with an accuracy of 96.111 %. The proposed technique implemented the best combination of laser wavelength and kernel function of 648 nm and RBF, respectively. Therefore, this innovative LLBI technique has the potential to optimize quality monitoring of horticultural products during postharvest handling, reduce fruit rejection rates, and prevent financial losses in the agro-industrial sector.