A Method of Spatial Mapping and Reclassification for High-Spatial-Resolution Remote Sensing Image Classification

Table 2

Producer and user classification accuracies for QuickBird image classification.

Class	Proposed method		Pixel-based SVM		Object-oriented SVM		SVMMV
	Producer's; accuracy %	User's; accuracy %	Producer's; accuracy %	User's; accuracy %	Producer's; accuracy %	User's; accuracy %	Producer's; accuracy %	User's; accuracy %
Road	83.14	80.24	49.72	79.21	55.39	77.88	50.59	79.37
Building	74.93	72.98	64.11	40.80	57.74	55.72	65.96	48.25
Vegetation	88.51	90.07	78.46	93.71	84.48	92.29	83.56	92.56
Water	74.05	96.61	66.97	83.15	74.05	93.33	72.85	91.14
Bare land	57.05	55.71	63.02	35.63	64.93	30.82	64.42	36.03
Shadow	68.00	75.75	63.63	67.66	62.53	83.85	74.33	73.44
Artificial grass	94.78	50.00	91.74	42.71	93.91	34.84	92.61	52.85
Plastic track	91.00	68.16	94.50	77.14	95.50	71.00	96.50	83.19
Average accuracy	78.93	73.69	71.52	65.00	73.57	67.47	75.10	69.60
Overall accuracy	81.01		69.01		72.48		73.04