Discrimination of Sidewalk Surface Condition Based on Image Textures and Meteorological Information

Abstract

Slippery roads, especially during and after a heavy snow fall, may lead to accidents causing injuries and fatalities to vulnerable person such as the aged. In this context, it is important to keep pedestrian aware of sidewalk condition. This paper aims at proposing detection of several sidewalk conditions under different environment circumstance. At the front end, image and video processing is performed to separate background and foreground images. Background image features are extracted using several texture feature generators. In this study, factor analysis methods are employed to examine the pattern of correlations among variables, and to reduce data dimensionality. Finally, Artificial Neural Network is employed to discriminate sidewalk surface condition, i.e., dry, wet, or snow.