Rice is the most important staple food across the world especially in Korean. Recently, rice gains popularity in US because it is one of the most healthy and gluten-free food material. In the view of rice processing, composition and characteristics of rice components such as carbohydrate, protein, and amylose are important. In order to produce commercial scale rice product s, automatic evaluation technology for the products in the processing is necessary. Hyperspectral imaging technology can acquire spectral and spatial information of the target materials at the same time so that it can provide an effective solution for the agro-food quality evaluation. Three kinds of flour samples, short white rice (Mimyeon, 75.6㎛, 14%), wheat (strong flour, 36.6 ㎛, 13.3%), and tapioca (tapioca starch, 27.5㎛, 14%) were prepared in a sample container of 5 × 5 × 2 cm. Samples were measured and classified in real time using on-line hyperspectral imaging system (wavelength : 930 ~ 1,700 nm, image : 320 pixels) with 60 Hz conveying speed. Colors in CIELAB color space of the each sample were measured using colorimeter. CIE L*a*b* values of wheat flour were 95.6±0.14, -1.00±0.04, 9.36±0.04, rice flour were 97.8±0.12, -1.02± 0.02, 3.48±0.04, and tapioca flour were 98.7±0.09, -0.27±0.04, 2.5±0.12, respectively. Based on these color informations, we developed an on-line model for the flour identification using hyperspectral imaging system. The model and the hyperspectral imaging system showed great potential for classifying three different types of flours.