Topological methods for structural properties in artificial vision

摘要

The generation of symbolic representations from particular data provided by sensors is a difficult problem. To specify this generation we must extract features, perform their grouping, select models and display rules to match with abstract models represented as equivalence classes certain topological transformations. Thus, a reasonable strategy to obtain closed-form solutions would consist in the acquisition of rough models associated to equilibrium solutions in the static case; the application of perturbation techniques to initial systems and solutions, will allow us from low-level to high-level models. In the way, we must replace equilibrium by stable solutions.