Vision-Based Satellite Recognition and Pose Estimation Using Gaussian Process Regression

Table 6

Pose estimation results on the 2D subset of BUAA-SID 1.5. Results are in bold if the proposed method performs the best and in italic if the second best.

Representation	Pose	Method	MAE (°)	(%)	(%)	(%)
BI	Pitch	MCGPR	27.8	8.23	14.01	28.99
		GPR	32.3	4.90	9.60	19.10
		Kernel method [19]	27.75	8.23	13.91	28.89
BI	Yaw	MCGPR	24.0	7.35	14.79	30.17
		GPR	30.6	3.92	7.25	16.85
		Kernel method [19]	24.04	7.35	14.89	29.97
DT	Pitch	MCGPR	15.4	8.03	15.67	33.69
		GPR	21.1	4.80	9.50	22.53
		Kernel method [19]	15.53	7.54	15.38	32.71
DT	Yaw	MCGPR	14.7	10.19	19.00	41.33
		GPR	23.6	4.31	9.01	22.92
		Kernel method [19]	14.85	10.97	19.10	40.84
HU	Pitch	MCGPR	58.7	1.08	2.45	6.17
		GPR	42.4	5.48	5.88	5.88
		Kernel method [19]	61.68	1.08	2.84	6.17
HU	Yaw	MCGPR	89.3	0.88	1.47	3.23
		GPR	90.0	0.29	0.39	0.98
		Kernel method [19]	89.34	0.69	1.57	3.33
FD	Pitch	MCGPR	40.8	1.86	4.11	10.48
		GPR	40.3	1.67	3.04	8.81
		Kernel method [19]	69.62	1.57	2.25	6.37
FD	Yaw	MCGPR	50.6	0.69	1.37	4.70
		GPR	52.5	0.69	0.98	3.82
		Kernel method [19]	75.95	0.59	1.47	5.39
GI	Pitch	MCGPR	1.04	69.05	87.07	98.04
		GPR	1.89	44.66	66.01	92.95
		Kernel method [19]	1.04	69.15	87.07	98.04
		ResNet [46]	3.35	19.5	37.81	74.93
		DenseNet [47]	3.84	12.73	35.65	72.87
GI	Yaw	MCGPR	0.94	72.87	88.05	97.65
		GPR	8.21	19.49	34.48	64.15
		Kernel method [19]	0.93	72.77	87.86	97.75
		ResNet [46]	47.62	2.25	4.21	8.42
		DenseNet [47]	47.45	1.18	2.64	5.29
HOG	Pitch	MCGPR	3.79	26.44	47.40	77.08
		GPR	7.99	13.52	26.93	55.53
		Kernel method [19]	3.79	26.44	47.31	77.08
HOG	Yaw	MCGPR	4.29	27.13	47.11	81.29
		GPR	13.7	7.74	14.79	36.14
		Kernel method [19]	4.29	27.23	47.01	81.29