2-D and 3-D touching part recognition using the theory of evidence

A unified approach is presented for solving the 2-D and 3-D touching part recognition problem. The problem is formulated as a Dempster-Shafer evidence accumulation process. Input line segment pairs of a scene (image-edge-pairs) are considered as evidence. This evidence is matched against line segment pairs of models (model-edge-pairs) to obtain object viewing parameters. The degree of support each piece of evidence provides to a model is calculated based on the viewing parameters. Using the theory of evidence all the different supports are accumulated. A decision about the identity of the object(s) in the scene is obtained from the combined evidence. Examples and experiments are conducted to confirm the robustness of the approach.<
>