The applications of robust estimation method BaySAC in indoor point cloud processing

Figures & data

Figure 1. 2D hypothesis model parameters histogram of a plane.

Figure 2. Flowchart of BaySAC-CONV algorithm.

Figure 3. Identified 3D correspondences: (a) Correspondences in scan A; (b) Correspondences in scan B.

Figure 4. Illustration of check points.

Table 1. Evaluation of registration accuracy in terms of average point-to-point distance.

Point cloud	Number of Check Points	RMS (RANSAC)	RMS (BaySAC-CONV)
Data-set I	5	50.2 mm	23.3 mm

Table 2. Computational efficiencies of the proposed strategies (Iteration).

Point cloud	Possible Correspondences	I/C ratio	Iterations (RANSAC)	Iterations (BaySAC-CONV)
				Random	Total
Data-set I	48	81%	6–18	1–6	2–11

Table 3. Computational efficiencies of the proposed strategies (unit: ms).

Point cloud	Time of RANSAC	Time of BaySAC-CONV
		Random	Total
Data-set I	21	3	5

Figure 5. Comparison of the fitting results of different methods on Data-set II: (a) RANSAC; (b) BaySAC-CONV.

Figure 6. Comparison of the fitting results of different methods before and after the point cloud simplification of Data-set III: (a) RANSAC; (b) BaySAC-CONV.

Table 4. Comparison of fitting results.

Point cloud	Method	Number of fitting planes	Average median error (m)	Fitting time (ms)
Data-set II	RANSAC	2707	0.028	3889
	BaySAC-CONV	1943	0.028	2540
Data-set III	RANSAC	48	0.0049	1492
	BaySAC-CONV	48	0.0074	337