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

Table 6

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

	KNN		SVM		Our approach
	Precision	Recall	Precision	Recall	Precision	Recall
Buildings	0.7556	1	0.8395	1	0.9527	1
Parking lot	0.5208	0.5855	0.7588	0.7295	0.8405	0.82077
Shrub	0.573	0.7801	0.8041	0.7705	0.9133	0.8582
Grass	0.5264	0.7266	0.8315	0.8853	0.9515	0.9474
Bare soil	0.6975	0.8316	0.791	0.7655	0.9097	0.8711
Road	0.701	0.7961	0.7229	0.7383	0.8836	0.8642

buildings are extracted separately from other types of object.