Recognition of object orientation from images

This paper proposes a new method for detection of orientation of a small vehicle from eight different orientations (e.g. front, front-left, front-right, side-left, side-right, back, back-left and back-right) of the vehicle with respect to a stationery camera. The proposed method is based on extracting features of a car object in terms of boundary description known as the signature. Matching of the signature with that of the templates stored in a database is performed towards deriving a dissimilarity value. This is used as a metric for classification. The proposed methodology may also be used for identifying the orientations of objects other than the cars.