EvRSO-FCSP: A Knowledge-Driven Framework for Complex Object Recognition Using Evolving Ontology and Fuzzy Constraint Reasoning

Knowledge-driven reasoning systems require structured representations capable of evolving over time while handling uncertainty, heterogeneous knowledge, and complex relational constraints. Although ontologies provide explicit semantic modeling and constraint-based methods enable rigorous reasoning, existing approaches generally rely on static knowledge representations and crisp constraint satisfaction, limiting their adaptability in dynamic and uncertain environments. To address these limitations, this paper proposes EVRSO-FCSP, a hybrid knowledge-driven reasoning framework that combines evolving ontologies with fuzzy constraint satisfaction for complex object recognition under uncertainty. The approach models objects through an evolving ontology that explicitly represents semantic concepts, spatial relationships, and inference rules, and automatically translates this knowledge into a fuzzy constraint network. A hybrid reasoning strategy integrating fuzzy arc-consistency, Particle Swarm Optimization, and adaptive backjumping is introduced to efficiently explore large solution spaces while preserving semantic consistency. The proposed framework is validated on the semantic interpretation of high-resolution remote sensing scenes involving complex geographic objects. Experimental results demonstrate improvements in recognition accuracy, spatial consistency, adaptability to previously unseen configurations, and reasoning robustness compared with representative knowledge-driven and learning-based baselines. These results demonstrate that combining ontology evolution with fuzzy constraint reasoning provides an effective and transferable paradigm for knowledge-based spatial reasoning under uncertainty.