Recognition of objects in various situations from two dimensional images

Generally, computers can successfully achieve object recognition by relying on sufficient information of observed objects. However, in real world, many objects own diverse configurations or objects are observed at various angles and positions, which make it difficult to match the observed objects with data models in a limited database. In this paper, to resolve the above problem, we propose an algorithm to achieve this kind of object recognition in shape space. Firstly, we describe the algorithm of extracting landmarks from outer contour of a shape by using recursive landmarks determination, in which the number of the landmarks can be appointed. Then, for the objects with many configurations, a series of new data are generated from one or two data models in pre-shape space. Finally, we achieve object recognition with shape space theory. The proposed method is efficient not only for the objects with noises, but also for the ones with various situations.