Decision-Tree Based Two-Dimensional Object Recognition
Rajiv Mehrotra, W.I. Grosky, F.K. Kung · 2005
A new model-based approach to two-dimensional object recoqnition is presented. Each known object is represented as a composition of its meaningful components. Some key components are selected from each model and these selected key components are stored in a binary decision-tree. An unknown object is recognized by selecting a key component from the scene representation and searching the decision-tree to obtain its possible identitie. Each possible identity of a key component of the scene is utilized in hypothesizing the identity and location of the posiible objects. The hypotheses are then tested for their valididty. The proposed approach is capable of recognizing both fully visible and partially occluded objects.