An Ontology-Based Framework for Complex Urban Object Recognition through Integrating Visual Features and Interpretable Semantics

Table 4

Quantitative evaluation of recognition results based on the first experimental datasets.

	KNN		SVM		Our approach
	Precision	Recall	Precision	Recall	Precision	Recall
Residential buildings	0.5852	0.6132	0.6877	0.6523	0.9688	0.9798
Commercial buildings	0.6721	0.5547	0.7621	0.6788	0.9167	0.9136
Circular roads	0.3894	0.6998	0.5586	0.5919	0.9705	0.9814
Swimming pool	0.702	0.3485	0.8775	0.9262	0.9333	1
Trees (1.5–5m)	0.5264	0.4416	0.7295	0.8562	0.8824	0.9525
Trees (5–10 m)	0.4799	0.5230	0.7462	0.8025	0.8493	0.9729
Trees (>10 m)	0.5012	0.4611	0.7119	0.837	0.8677	0.9702