Reasoning about the recognition of uncertain objects

Abstract

This thesis describes a vision expert system which recognizes the input objects through the inference process for deformed object. The system analyzes the input object by structural description and classifies the object into similar structural set in hierarchical object knowledge base (HOKB). Next the system matches the object to model objects in terminal class of HOKB and infers a deformed or missing part of the object in the domain of tools due to shadow, reflection and unreliable low level processing. The certainty factor (CF) model in hierarchical hypothesis space is applied to inference procedure under the deformation of objects.