Object recognition via hierarchical syntactic models
Basilis Gidas, Antonio Zelic · 2002
We propose a multiple-object recognition framework based on: (a) context-free-grammars type hierarchical syntactic models for representing objects in a database, and (b) nonparametric statistics (rank tests of Kolmogorov-Smirnov statistics) for designing local descriptors of grey-level image data. The procedure has been successfully tested on a database of 2-D simulated tools in an environment highly degraded by noise, blur, clutter, and occlusion.