Fuzzy relations for feature-model correspondence in 3D object recognition

This paper presents a new mechanism for determining feature correspondences for object recognition, based on fuzzy set theory. The new method applies unary and binary constraints from the model, taking uncertainty characteristics of the measurement process into consideration. Experiments with both unoccluded and occluded images show that the method selects an appropriate set of correspondences, especially when integrated with fuzzy perceptual grouping.