A fuzzy approach to pose determination in object recognition

Object recognition is the process of identifying and locating known objects in complex images. It includes extracting relevant features, grouping these features together, selecting an appropriate object model, and determining the pose (position and orientation) of the object in the scene. In earlier work, the author has shown that fuzzy methods are appropriate for representing geometric relationships that are used for both perceptual grouping of geometric features and for associating geometric image features with models. The paper explores fuzzy methods for the final step in object recognition, that of global pose determination. She develops a method based on fuzzy c means (FCM) clustering, and demonstrates its effectiveness over traditional crisp pose clustering.