Object recognition from 3D blurred images

摘要

This paper presents a novel approach for 3D object recognition and classification. The originality of this approach is the combination of statistical reasoning, feature learning and expertise knowledge from the application. It is based on 2D processing of images alternating with 3D validation of class characteristics for recognition. Classification is based on iterative refinement of hypotheses. An application is presented for pollen grain observed with light microscopy for allergy prevention.