Advancing mango quality assurance: Non‐destructive detection of spongy tissue using visible near‐infrared spectroscopy and machine learning classification

Abstract

The renowned Alphonso mango, celebrated in India for its exquisite taste, saffron hue, pleasing texture, and prolonged shelf life, holds significant global commercial appeal. Unfortunately, the widespread issue of spongy tissue (ST) disorder in Alphonso mangoes results in a soft, corky texture, affecting up to 30% of mangoes in a single batch. This challenge leads to the outright rejection of affected mangoes during export due to delayed disorder identification. The current evaluation method involves destructive cutting, causing substantial fruit loss, and lacks assurance for the overall batch quality. This study addresses these challenges by focusing on the utilization of visible near‐infrared spectroscopy as a non‐invasive method to assess the internal quality of mangoes. Additionally, it introduces innovative classification models for automated binary categorization (Healthy vs. ST affected). Through preprocessing and principal component analysis of spectral reflectance data for feature extraction and wavelength optimization, successful wavelength ranges of 650–970 nm were identified, effectively distinguishing between healthy and damaged mangoes. Various machine learning models used notably, linear discriminant analysis, support vector machine, and logistic regression exhibited strong discriminative capabilities with higher accuracy reaching 99%. This non‐destructive approach addresses critical challenges in the mango export industry, offering early detection of internal disorders and minimizing postharvest losses.