Fine-Grained Point Cloud Semantic Segmentation of Complex Railway Bridge Scenes from UAVs Using Improved DGCNN

Table 5

Results of segmentation of three scenes in different areas.

Scenes	Test area	IoU (%)							mIoU (%)	bACC (%)
		Cable	Cluster	Guardrail	Mast	Protective wall	Rail	Track bed
(a)	1	99.46	99.56	99.92	97.88	99.08	97.76	99.21	98.98	99.51
	2	99.61	99.71	99.96	98.43	99.51	98.37	99.32	99.27	99.68
	3	99.40	99.82	99.99	97.86	99.60	98.53	99.52	99.24	99.68
	4	99.10	99.58	99.95	97.90	99.42	99.10	99.45	99.21	99.63
	Mean	99.39	99.66	99.88	98.02	99.40	98.44	99.38	99.18	99.63
(b)	1	98.74	90.27	97.80	91.07	92.92	83.80	86.77	91.62	96.73
	2	98.44	88.31	—	92.97	—	82.83	87.63	90.04	95.15
	3	98.78	87.84	97.12	90.20	91.70	82.05	86.82	90.64	95.31
	4	98.56	88.49	—	92.42	—	84.80	85.86	90.03	95.07
	5	98.77	89.63	98.62	90.41	91.96	83.84	87.18	91.49	96.66
	Mean	98.66	88.91	97.85	91.41	92.19	83.46	86.85	90.76	95.78
(c)	1	97.50	84.86	—	89.45	—	74.29	—	86.53	92.11
	2	96.95	83.45	—	88.23	—	75.45	82.20	85.26	91.05
	3	97.04	83.39	—	88.91	—	75.84	—	86.30	91.77
	4	97.84	82.84	94.46	89.59	85.99	72.54	—	87.21	92.64
	5	96.38	81.55	93.80	86.76	83.95	73.53	82.64	85.52	91.26
	6	—	83.70	96.95	88.70	78.84	72.94	—	84.23	90.87
	Mean	97.14	83.30	95.07	88.61	82.93	74.10	82.42	85.84	91.62