Hyperspectral imaging systems have been recently popular for food quality and safety assessment, due to their ability to reflect unique spectral properties of materials at narrow band intervals. They help detection of contamination in foods such as dash, mold, crush, fungi. We propose such a system for effective detection of black molds in dried figs to avoid the high cost of manual process. The proposed system depends on finding the best discriminative spectral band and the optimum local spatial characteristics using forward feature selection with commonly used classifiers. The preliminary accuracies of upto 93.58% are promising for an operational system.