OBJECT RECOGNITION WITH CONSTRAINED ELASTIC MODELS

We present a model-based method for object identification in images of natural scenes. It has successfully been implemented for the classifi cation of cars based on their rear view. In a first step, characterist ic features such as lines and corners are detected within the image. G eneric models of object-classes, described by the same set of features , are stored in a database. Each model represents a whole class of obj ects (e.g., passenger cars, vans, big trucks). In a preprocessing stag e, the most probable object is selected by means of a corner-feature b ased Hough transform. This transformation also suggests the position a nd scale of the object in the image. Guided by similarity measures, th e model is then aligned with image features using a matching algorithm based on the elastic net technique [1]. During this iterative process , the model is allowed to undergo changes in scale, position and certa in deformations. Deformations are kept within limits such that one mod el can fit to all objects belonging to the same class, but not to obje cts of other classes. In each iteration step, quantities to assess the matching process are obtained.