Enhanced Point Cloud Segmentation Method Using Positional Encoding and Channel Attention

Wei Zhang; Zhilong Zeng; Qi Fang; Jie Song; Guan Gui; Shenghuai Wang; Chen Wang

doi:10.3788/LOP241294

Laser & Optoelectronics Progress, Volume. 62, Issue 4, 0415008(2025)

Enhanced Point Cloud Segmentation Method Using Positional Encoding and Channel Attention

Wei Zhang^1,2、*, Zhilong Zeng¹, Qi Fang¹, Jie Song¹, Guan Gui¹, Shenghuai Wang^1,2, and Chen Wang^1,2

Author Affiliations

¹College of Mechanical Engineering, Hubei University of Automotive Technology, Shiyan 442002, Hubei , China

²Hubei Zhongcheng Technology Industry Technique Academy Co., Ltd., Shiyan 442003, Hubei , China

show less

Abstract Get PDF(in Chinese)

Figures & Tables(12)

Fig. 1. PCANet structure

Download full size

Fig. 2. Relative position encoding module

Download full size

Fig. 3. CAA module

Download full size

Fig. 4. Compact channel-wise comparator module and channel affinity estimator module

Download full size

Fig. 5. Visualization of the segmentation results in ShapeNet. (a) Segmentation truth graph; (b) DeLA segmentation results; (c) PCANet segmentation results

Download full size

Fig. 6. Visualization of the segmentation results in S3DIS. (a) Input point clouds; (b) segmentation results of SPoTr; (c) segmentation results of PCANet

Download full size

Table 1. Part segmentation results on the ShapeNet dataset

View table

Table 1. Part segmentation results on the ShapeNet dataset

Method	mean	aero	bag	cap	car	chair	earphone	quitar	knife	lamp	laptop	motor	mug	pistol	rocket	skateboard	table
PointNet	83.7	83.4	78.7	82.5	74.9	89.6	73.0	91.5	85.9	80.8	95.3	65.2	93.0	81.2	57.9	72.8	80.6
PointNet++	85.1	82.4	79.0	87.7	77.3	90.8	71.8	91.0	85.9	83.7	95.3	71.6	94.1	81.3	58.7	76.4	82.6
SCN^［24］	84.6	83.8	80.8	83.5	79.3	90.5	69.8	91.7	86.5	82.9	96.0	69.2	93.8	82.5	62.9	74.4	80.8
DGCNN	85.2	84.0	83.4	86.7	77.8	90.6	74.7	91.2	87.5	82.8	95.7	66.3	94.9	81.1	63.5	74.5	82.6
3D-GCN	85.1	83.1	84.0	86.6	77.5	90.3	74.1	90.9	86.4	83.8	95.6	66.8	94.8	81.3	59.6	75.7	82.8
PointMLP^［25］	86.1	83.5	83.4	87.5	80.5	90.3	78.2	92.2	88.1	82.6	96.2	77.5	95.8	85.4	64.6	83.3	84.3
SPoTr^［26］	87.2
DeLA^［27］	87.0	86.9	86.3	88.4	83.5	92.5	78.6	92.9	87.8	84.8	96.7	79.4	96.6	84.5	65.7	78.4	83.8
PCANet	87.4	86.7	88.4	89.0	83.1	92.3	79.9	92.5	87.8	85.7	96.8	79.8	96.8	84.5	66.4	78.3	83.9

Table 2. Part segmentation Cat. mIoU and Ins. mIoU results on the ShapeNet dataset
View table
Table 2. Part segmentation Cat. mIoU and Ins. mIoU results on the ShapeNet dataset
Method Cat. mIoU Ins. mIoU
PointNet 80.4 83.7
PointNet++ 81.9 85.1
DGCNN 82.3 85.2
PointMLP 84.6 86.1
SPoTr 85.4 87.2
PCANet 85.8 87.4

Table 3. Results of instances with different K values
View table
Table 3. Results of instances with different K values
num-points K Ins. mIoU /%
2048 0 85.42
2048 10 85.62
2048 15 85.64
2048 20 85.75
2048 25 85.50
2048 30 85.60

Table 4. Module importance analysis
View table
Table 4. Module importance analysis
Position Encode CCC CAE Ins. mIoU /%
85.0
√ 85.4
√ √ 85.5
√ √ √ 85.8

Table 5. Semantic segmentation results of S3DIS
View table
Table 5. Semantic segmentation results of S3DIS
Method OA mAcc mIoU
PointNet 41.1
PointNet++ 53.5
DGCNN 84.1 56.1
KPConv^［18］ 72.8 67.1
PointNeXt^［28］ 91.0 77.2 70.5
SPoTr 90.7 76.4 70.8
PTv3^［29］ 91.8 79.4 73.4
PCANet 92.1 79.9 73.7

Table 6. Semantic segmentation time of S3DIS
View table
Table 6. Semantic segmentation time of S3DIS
Method Every epoch /s
PointNet 104
PointNet++ 140
DGCNN 225
PointNeXt 318
SPoTr 273
PCANet 148

Tools

Get Citation

Copy Citation Text

Wei Zhang, Zhilong Zeng, Qi Fang, Jie Song, Guan Gui, Shenghuai Wang, Chen Wang. Enhanced Point Cloud Segmentation Method Using Positional Encoding and Channel Attention[J]. Laser & Optoelectronics Progress, 2025, 62(4): 0415008

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites