Classification of black mold contaminated figs by hyperspectral imaging

With increased demands and expectations for food quality, the agricultural industry has to provide non-destructive, safe, fast and reliable methods to meet these challenges. Thus, in recent years, hyperspectral imaging systems have gained increasing importance for food quality assessment in various applications including poultry, meat, vegetables and fruits. We propose such a system for automated evaluation of dried figs. Dried figs, which are economically important for rural development, are easily affected by black mold during their process. Traditional way to detect the black mold contaminated figs, which depends on human inspection, is labour expensive, time consuming and it carries the risks of transmitting the molds to the sound figs. Our proposed system based on hyperspectral image analysis eliminates these disadvantages and provides effective discrimination of the black-mold contaminated figs. The system acquires images of figs at 784 spectral bands, and selects the most discriminative 50 bands using the sequential floating forward selection, and then extracts their spatial characteristics to localize the contaminated region. The high accuracies obtained for initial small dataset with representative samples are promising for an operational system.