Visual Map Construction Using RGB-D Sensors for Image-Based Localization in Indoor Environments

Table 2

Position errors of the database camera with different mapping methods.

Scenes	Methods of modeling	Average errors (meters)						Accuracy improvement (%)
		In direction		In direction		Of Euclidean distance
		Day	Night	Day	Night	Day	Night	Day	Night
Small room	Proposed method	0.0565	0.0440	0.0424	0.0397	0.0750	0.0640	—	—
	RGB-D ICP	0.0739	0.0721	0.0617	0.0566	0.1040	0.0980	27.88	34.69
	Classical ICP	0.0822	0.0823	0.0726	0.0759	0.1186	0.1187	36.76	46.08
Medium room	Proposed method	0.0608	0.0592	0.0418	0.0296	0.0775	0.0695	—	—
	RGB-D ICP	0.0906	0.0874	0.0761	0.0691	0.1237	0.1197	37.35	41.94
	Classical ICP	0.1211	0.1220	0.0902	0.0883	0.1629	0.1632	52.42	57.41
Large room	Proposed method	0.1187	0.1089	0.0975	0.0903	0.1643	0.1525	—	—
	RGB-D ICP	0.2318	0.2374	0.1946	0.1702	0.3247	0.3144	49.40	51.49
	Classical ICP	0.3578	0.3171	0.2241	0.2029	0.4539	0.4505	63.80	66.15
Corridor	Proposed method	0.0800	0.0838	0.0413	0.0411	0.0953	0.0977	—	—
	RGB-D ICP	0.1584	0.1644	0.0537	0.0525	0.1727	0.1787	44.82	45.33
	Classical ICP	0.1970	0.2057	0.1253	0.1216	0.2518	0.2548	62.15	61.66