Moving vehicle shadow elimination approach based on mark growing of multi-feature fusion

摘要

In order to eliminate vehicle shadow in moving vehicle detection, a novel vehicle shadow elimination approach is presented based on mark growing of multi-feature fusion. The selection of mark points and the establishment of growing rules are critical. Firstly, background model is obtained by mixture Gaussian method. Then gradient difference of foreground and background is calculated. Morphological method is adopted to recover approximate shadow area. Typical shadow points or mark points are got from this area. As shadow area is less bright than object area and the color inside shadow area is similar, the growing rules are established. Growing is carried out starting from mark points based on growing rules to eliminate shadows. As combining various features such as texture, brightness and color, the experiment results prove the effectiveness of this method.