We introduce in this paper a shadow detection method for vehicles in traffic video sequences. Our method approximates the boundary between vehicles and their associated shadows by one or more straight lines. These lines are located in the image by exploiting both local information (e.g. statistics in intensity differences) and global information (e.g. principal edge directions). The proposed method does not assume a particular lighting condition, and requires no human interaction nor parameter training. Experiments on practical real-world traffic video sequences demonstrate that our method is simple, robust and efficient under traffic scenes with different lighting conditions. Accurate positioning of target vehicles is thus achieved even in the presence of cast shadows.
展开▼
