Tomato quality assessment and enhancement through Fuzzy Logic: A ripe perspective on precision agriculture

Abstract

This study presents a meticulous exploration of tomato quality assessment and enhancement employing Fuzzy Logic, focusing on beefsteak, Roma, and cherry varieties. The research incorporates a multivariate approach encompassing shape, size, texture, and color analysis. Utilizing a controlled imaging system, 165 tomatoes were captured at various ripening stages to ensure uniformity in lighting, camera position, and background color. Image processing and computer vision techniques, including segmentation and feature extraction, were applied to comprehensively evaluate the tomatoes. Quality assessment involved considerations of three stages of skin coloration, defect quantity, and skin texture calculated using the statistical method of entropy. The decision-making process and quality evaluation were facilitated by a Mamdani-type Fuzzy Inference System, leveraging numerical values of external characteristics such as dimension, calyx shape, ripeness stage, and defect regions. The fuzzy logic-based system employed 22 fuzzy rules and three different membership functions (trapezoidal, Gaussian, and triangular), enabling a nuanced mapping of transitions between classes within the input variables. Furthermore, the system underwent rigorous training and validation using 80% and 20% respectively of the captured images. Through the system, we were able to identify and segment areas of tomatoes exhibiting poor conditions. This approach extends beyond conventional methods of estimating fruit ripeness, which typically rely on color assessment or destructive firmness evaluations.