Key Contour Feature Extraction of Printed Circuit Board Point Cloud

Wenbin Zhong; Xurui Li; Si Sun; Guangshuai Liu

doi:10.3788/LOP57.141001

Journals >Laser & Optoelectronics Progress >Volume 57 >Issue 14 >Page 141001 > Article

Laser & Optoelectronics Progress
Vol. 57, Issue 14, 141001 (2020)

Key Contour Feature Extraction of Printed Circuit Board Point Cloud

Wenbin Zhong¹, Xurui Li^2、*, Si Sun², and Guangshuai Liu²

Author Affiliations

¹The 10th Research Institute of China Electronics Technology Group Corporation, Chengdu, Sichuan 610036, China

²School of Mechanical Engineering, Southwest Jiaotong University, Chengdu, Sichuan 610031, China

show less

DOI: 10.3788/LOP57.141001 Cite this Article Set citation alerts

Wenbin Zhong, Xurui Li, Si Sun, Guangshuai Liu. Key Contour Feature Extraction of Printed Circuit Board Point Cloud[J]. Laser & Optoelectronics Progress, 2020, 57(14): 141001 Copy Citation Text

show less

Fig. 1. Block diagram of key contour feature extraction algorithm of printed circuit board

Download full size

Fig. 2. Determination of the projection plane of the boundary point

Download full size

Fig. 3. Defining conditions of boundary points.(a)Point is on the boundary contour;(b)point is not on the boundary contour

Download full size

Fig. 4. Comparison of feature line extraction effects of 6 different algorithms on printed circuit board. (a) Raw point cloud; (b) r search algorithm; (c) k-NNS algorithm; (d) concave polygon algorithm; (e) convex hull algorithm; (f) RANSAC algorithm; (g) proposed algorithm

Download full size

Fig. 5. Comparison diagrams of the extraction effects. (a)(c) k-NNS algorithm; (b)(d) r search algorithm

Download full size

Fig. 6. Comparison diagrams of the extraction effects. (a)(c) RANSAC algorithm; (b)(d) proposed algorithm

Download full size

Fig. 7. Part contour size of the printed circuit board

Download full size

Input:Point cloud: P= $\{p_{1}, \dots, p_{n}\}$ Parameter: i,j,k,N_min,N_max,d_th,p_r,n,α_thVector: C,Q[j]
1: For i=0 to size $\{P\}$ do
2: If p_r[i]=true then3: continue
4: Q.push_back(i)5: p_r[i]=true6: End if7: For j=1 to j do8: set p_q=p_j as query point9: Use kd-tree to find the k nearest neighbor $p_{i}^{k}$
10: If L(p_q, $p_{i}^{k}$ )≤d_th & arccos (<n_q,n_k>)≤α_th then11: Add $p_{i}^{k}$ to Q[j]12: j=j+113: p_r[i]=true14: End if15:End for
16: If N_min≤Q $[size]$ ≤N_max then17: C.push_back Q[j]
18: clear Q19: i=i+120: End if21: End for22:Return C

Table 1. Euclidean clustering based on normal vector angle

Algorithm	Number of reservedfeature points	Time /s
r search	1658	178.0
k-NNS	4525	24.0
Concave polygon	281	4.0
Convex hull	183	2.0
RANSAC	6816	28.3
Proposed algorithm	7801	20.8

Table 2. Statistical results of number of reserved feature points after boundary point extraction

Algorithm module	Time /s
Pre-processing	0.096
Detection of maximum plane extraction	1.572
Remaining point extraction	1.672
Boundary 1 extraction	2.773
Boundary 2 extraction	0.474
B1∪B2	0.219
Total time	6.806

Table 3. Time-consuming of each sub-module of proposed algorithm

Key contour size	Measurement result /mm	Actual value /mm	Actual error /mm	Relative error /%
L1	148.502	148.600	-0.098	-0.065
L2	115.739	115.840	-0.101	-0.087
L3	16.320	16.200	0.120	0.740
L4	16.233	16.260	-0.027	-0.166
L5	14.921	14.880	0.041	0.275

Table 4. Comparison between measurement results of contour size and actual values

Wenbin Zhong, Xurui Li, Si Sun, Guangshuai Liu. Key Contour Feature Extraction of Printed Circuit Board Point Cloud[J]. Laser & Optoelectronics Progress, 2020, 57(14): 141001

Download Citation

Set citation alerts for the article

Tools

Set citation alerts for the article

Save the article for my favorites

Paper Information