| Face detection and eye detection have more advantages compared with other body feature detection, used extensively and distinguishingly in social security, and became research hotspot in pattern recognition, artificial intelligence and computer vision. We presented fast eye detection and drowsiness detection method for intelligent classroom using a common low resolution camera, which showed its robustness in various illumination, gesture and intricate background.In the first paragraph, we described background, objective, and siginificance for drowsiness detection. After analyzed drowsiness scenes in intelligent classroom and compared several technologies, we chose digital image processing and fuzzy logic, and the overall flow chart was given subsequently.The detection process was divided into two parts-face detection and eye detection. We gave the brief introduction of classical face detection technologies;Skin model showed its advantage for the application of fast and rough face detection or localization. Students'head and shoulder images are acquired through the combination of background difference and frames difference. Applied skin color segmentation based on YCb ' Cr ' color space, we got a rough face area.For eye detection, we had two steps. Firstly, we maximized the eye features, extracted by synthesizing eye luminance, chroma and RGB features with the use of mathematics morphologic processing. Thus we got the eye candidate blocks. Secondly, to filter the candidate blocks, we established a fuzzy logic system based on statistical data. Six features were selected as inputs of the system: location (x and y), size (width, height and width/height), and hardness for eclipse. The fuzzy logic system could not only judge but also infer. After filtering the candidates, we did the last verification-brow and eye matching, and eye symmetry, and then the ultimate eye images were intercepted, from which eye state were judged. |
PDF Full Text Request