Non-Destructive Quality Assessment Technologies for Fresh Horticultural Produce in Nigerian Market Contexts: Development and Validation

Non-destructive quality assessment technologies offer transformative potential for reducing post-harvest losses and improving market efficiency in horticultural value chains by enabling rapid, objective quality evaluation without product damage. However, the development and validation of such technologies under Nigerian market conditions, where produce diversity, handling variability, and infrastructure constraints differ markedly from temperate contexts, has received minimal research attention. This dissertation develops and validates handheld near-infrared spectroscopy (NIRS) and computer vision-based quality assessment protocols for tomato, mango, and citrus under Nigerian market conditions. Spectral data were collected from 2,400 fruit samples sourced from six geopolitical zones across varying quality grades and handling histories. Partial least squares regression and convolutional neural network models were trained, validated, and tested for prediction of total soluble solids, firmness, lycopene content, and decay risk. Model performance was evaluated against laboratory reference measurements and field expert assessments under simulated Nigerian market conditions including variable lighting, ambient temperature, and produce contamination. NIRS-based models achieved prediction accuracy (R2 greater than 0.92) for soluble solids and firmness in tomato and mango. The convolutional neural network achieved 91 percent accuracy in decay risk classification for citrus. The study provides validated non-destructive quality protocols adaptable for low-cost smartphone-based implementation in Nigerian produce markets. Keywords: non-destructive quality assessment, near-infrared spectroscopy, computer vision, post-harvest, Nigerian markets