A General Polygon-based Deformable Model for Object Recognition

We propose a general scheme for object localization and recognition based on a deformable model. The model combines shape and image properties by warping a arbitrary prototype intensity template according to the deformation in shape. The shape deformations are constrained by a probabilistic distribution, which combined with a match of the warped intensity template and the image form the final criteria used for localization and recognition of a given object. The chosen representation gives the model an ability to model an almost arbitrary object. Beside the actual model a full general scheme for applying the model is proposed. The scheme includes general methods for initialization, optimization and validation. Experimental results for real data are shown. Compared to related work the proposed meodel and the methods for initialization and validation contains a number of intersting features and improved abilities.