Point-MLPBLS： point cloud semantic segmentation network based on MLP cascaded broad learning system

doi:10.11772/j.issn.1001-9081.2025060788

Journal of Computer Applications ›› 2026, Vol. 46 ›› Issue (7): 2259-2266.DOI: 10.11772/j.issn.1001-9081.2025060788

• Multimedia computing and computer simulation • Previous Articles

Point-MLPBLS： point cloud semantic segmentation network based on MLP cascaded broad learning system

Guoyou ZHANG(), Hongyu NIE, Lihu PAN, Rundong LEI

School of Computer Science and Technology，Taiyuan University of Science and Technology，Taiyuan Shanxi 030024，China

Received:2025-07-16 Revised:2025-10-14 Accepted:2025-10-15 Online:2025-10-27 Published:2026-07-10
Contact: Guoyou ZHANG
About author:NIE Hongyu， born in 2000， M. S. candidate. His research interests include broad learning， intelligent software engineering.
PAN Lihu， born in 1974， Ph. D.， professor. His research interests include artificial intelligence， software engineering.
LEI Rundong， born in 2001， M. S. candidate. His research interests include intelligent software engineering， 3D recognition.
Supported by:
Natural Science Foundation of Shanxi Province(202403021221151)

基于多层感知机级联宽度学习系统的点云语义分割网络Point-MLPBLS

张国有(), 聂宏宇, 潘理虎, 雷润东

太原科技大学计算机科学与技术学院，太原 030024

通讯作者: 张国有
作者简介:聂宏宇（2000—），男，山东菏泽人，硕士研究生，主要研究方向：宽度学习、智能软件工程
潘理虎（1974—），男，河南上蔡人，教授，博士，CCF会员，主要研究方向：人工智能、软件工程
雷润东（2001—），男，陕西渭南人，硕士研究生，主要研究方向：智能软件工程、三维识别。
基金资助:
山西省自然科学基金资助项目(202403021221151);山西省自然科学基金资助项目(202203021221145)

Abstract

Abstract:

Semantic segmentation of 3D point clouds is a key technology in the fields of autonomous driving and computer vision. Deep learning-based semantic segmentation models suffer from inefficient operation due to their complex network structures. To address the inefficiency of deep neural networks in 3D point cloud semantic segmentation， Point-MLPBLS， a point cloud semantic segmentation network based on Multi-Layer Perceptron （MLP） cascaded Broad Learning System （BLS） was proposed. First， the overall network adopted a dual network collaborative architecture： a feature extraction network and a point cloud segmentation network. Second， the feature extraction network performed a sampling-grouping-aggregation operation to the input point cloud data， and an inverted residual module was added to enhance the local feature representation capability. Third， a feature propagation module was used to achieve upsampling， thereby restoring the original point cloud resolution and enabling multi-scale feature fusion， thereby constructing a refined point cloud representation. Finally， the point cloud segmentation network replaced deep iterative training with a flattened MLP cascade structure， and spatially-aware MLPs were embedded in the feature mapping layer， so as to achieve high-precision inference. Experimental results show that Point-MLPBLS improves the mean Intersection-over-Union （mIoU） by 12.9 percentage points and reduces the segmentation time by 41.1% compared with PointNet++ on the S3DIS dataset， improving the efficiency of point cloud segmentation， and providing an efficient solution for 3D point cloud semantic segmentation.

Key words: 3D point cloud, semantic segmentation, Broad Learning System (BLS), feature fusion, attention mechanism

摘要：

三维点云语义分割是自动驾驶与计算机视觉领域的关键技术。基于深度学习的语义分割模型由于网络结构复杂，运行效率低下。针对深度神经网络在三维点云语义分割中效率低下的问题，提出一种基于多层感知机（MLP）级联宽度学习系统（BLS）的点云语义分割网络Point-MLPBLS。首先，整体网络采用双网络协同架构：特征提取网络和点云分割网络；其次，特征提取网络对输入的点云数据采用采样-分组-聚合操作，并追加倒置残差模块增强局部特征的表达能力；再次，通过特征传播模块实现上采样，恢复原始点云分辨率并进行多尺度特征融合，构建精细化的点云表征；最后，点云分割网络采用扁平化MLP级联结构替代深层迭代训练，在特征映射层嵌入空间感知型MLP，实现高精度推理。在S3DIS数据集上的实验结果表明：与PointNet++相比，Point-MLPBLS的平均交并比（mIoU）提高了12.9个百分点，分割时间缩短了41.1%，提高了点云分割的效率，为三维点云语义分割提供了高效解决方案。

关键词: 三维点云, 语义分割, 宽度学习系统, 特征融合, 注意力机制

CLC Number:

TP391.41

Guoyou ZHANG, Hongyu NIE, Lihu PAN, Rundong LEI. Point-MLPBLS： point cloud semantic segmentation network based on MLP cascaded broad learning system[J]. Journal of Computer Applications, 2026, 46(7): 2259-2266.

张国有, 聂宏宇, 潘理虎, 雷润东. 基于多层感知机级联宽度学习系统的点云语义分割网络Point-MLPBLS[J]. 《计算机应用》唯一官方网站, 2026, 46(7): 2259-2266.

Figures/Tables 10

References 19

[1]	Wang J， Liu Y， Tan H， et al. A survey on weakly supervised 3D point cloud semantic segmentation［J］. IET Computer Vision， 2024， 18（3）： 329-342.
[2]	Mi X， Yang B， Chen C， et al. Automatic road structure detection and vectorization using MLS point clouds ［J］. The International Archives of the Photogrammetry， Remote Sensing and Spatial Information Sciences， 2019， XLII-2/W12： 1069-1075.
[3]	Qin H， Shao S， Wang T， et al. Review of autonomous path planning algorithms for mobile robots ［J］. Drones， 2023， 7（3）： No.211.
[4]	Tan C S， Mohd-Mokhtar R， Arshad M R. A comprehensive review of coverage path planning in robotics using classical and heuristic algorithms［J］. IEEE Access， 2021， 9： 119310-119342.
[5]	饶加旺，周松，邢策梅，等.城市三维模型通用格式转换与可视化研究［J］.计算机科学与应用， 2023， 13（4）： 903-914.
	Rao Jiawang， Zhou Song， Xing Cemei， et al. Research on universal format conversion and visualization of urban 3D model ［J］. Computer Science and Applications， 2023， 13（4）： 903-914.
[6]	王明军，李乐，易芳，等.模拟真实水体环境下目标激光点云数据的三维重建与分析［J］.中国激光， 2022， 49（3）： No.0309001.
	Wang Mingjun， Li Le， Yi Fang， et al. Three-dimensional reconstruction and analysis of target laser point cloud data in simulated real water environment ［J］. Chinese Journal of Lasers， 2022， 49（3）： No.0309001.
[7]	王涛，王文举，蔡宇.基于深度学习的三维点云语义分割方法研究［J］.计算机工程与应用， 2021， 57（23）： 18-26.
	Wang Tao， Wang Wenju， Cai Yu. Research of deep learning-based semantic segmentation for 3D point cloud ［J］. Computer Engineering and Applications， 2021， 57（23）： 18-26.
[8]	Qi C R， Su H， Mo K C， et al. PointNet： deep learning on point sets for 3D classification and segmentation ［C］// CVPR 2017. Piscataway： IEEE， 2017： 77-85.
[9]	Qi C R， Yi L， Su H， et al. PointNet++： deep hierarchical feature learning on point sets in a metric space ［C］// NeurIPS 2017. Red Hook： Curran Associates Inc.， 2017： 5105-5114.
[10]	Huang Q， Wang W， Neumann U. Recurrent slice networks for 3D segmentation of point clouds ［C］// CVPR 2018. Piscataway： IEEE， 2018： 2626-2635.
[11]	Wang Y， Sun Y， Liu Z， et al. Dynamic graph CNN for learning on point clouds ［J］. ACM Transactions on Graphics， 2019， 38（5）： No.146.
[12]	Hua B S， Tran M K， Yeung S K. Pointwise convolutional neural networks ［C］// CVPR 2018. Piscataway： IEEE， 2018： 984-993.
[13]	Simonovsky M， Komodakis N. Dynamic edge-conditioned filters in convolutional neural networks on graphs ［C］// CVPR 2017. Piscataway： IEEE， 2017： 29-38.
[14]	Ma Y， Guo Y， Liu H， et al. Global context reasoning for semantic segmentation of 3D point clouds ［C］// WACV 2020. Piscataway： IEEE， 2020： 2920-2929.
[15]	Park C， Jeong Y， Cho M， et al. Fast point Transformer ［C］// CVPR 2022. Piscataway： IEEE， 2022： 16928-16937.
[16]	Chen C L P， Liu Z. Broad learning system： an effective and efficient incremental learning system without the need for deep architecture ［J］. IEEE Transactions on Neural Networks and Learning Systems， 2018， 29（1）： 10-24.
[17]	Armeni I， Sener O， Zamir A R， et al. 3D semantic parsing of large-scale indoor spaces ［C］// CVPR 2016. Piscataway： IEEE， 2016： 1534-1543.
[18]	Li Y， Bu R， Sun M， et al. PointCNN： convolution on X-Transformed points ［C］// NeurIPS 2018. Red Hook： Curran Associates Inc.， 2018： 828-838.
[19]	Qian G， Li Y， Peng H， et al. PointNeXt： revisiting PointNet++ with improved training and scaling strategies ［C］// NeurIPS 2022. Red Hook： Curran Associates Inc.， 2022： 23192-23204.

N_f	N_e	OA/%	mIoU/%
100	5 000	87.7	67.2
100	10 000	87.9	67.4
200	10 000	88.3	67.7
200	12 000	88.6	68.0
400	10 000	88.8	68.3
400	11 000	88.5	68.1
800	10 000	88.2	67.8
800	12 000	87.9	67.7
1 000	10 000	87.7	67.4
1 000	13 000	87.5	67.4
2 000	12 000	87.4	67.2

N_f	N_e	OA/%	mIoU/%
100	5 000	87.7	67.2
100	10 000	87.9	67.4
200	10 000	88.3	67.7
200	12 000	88.6	68.0
400	10 000	88.8	68.3
400	11 000	88.5	68.1
800	10 000	88.2	67.8
800	12 000	87.9	67.7
1 000	10 000	87.7	67.4
1 000	13 000	87.5	67.4
2 000	12 000	87.4	67.2

网络	mIoU/%	OA/%	mAcc/%	分割时间/min
PointNet^［8］	42.3	80.4	62.1	46.2
PointNet++^［9］	55.2	82.6	68.8	53.8
DGCNN^［11］	48.6	82.8	70.2	49.6
PointCNN^［18］	62.6	83.9	72.3	50.2
Point Transformer^［15］	69.6	91.1	79.7	61.7
PointNeXt^［19］	67.2	87.3	76.4	55.3
Point-MLPBLS	68.1	88.7	79.1	31.7

网络	mIoU/%	OA/%	mAcc/%	分割时间/min
PointNet^［8］	42.3	80.4	62.1	46.2
PointNet++^［9］	55.2	82.6	68.8	53.8
DGCNN^［11］	48.6	82.8	70.2	49.6
PointCNN^［18］	62.6	83.9	72.3	50.2
Point Transformer^［15］	69.6	91.1	79.7	61.7
PointNeXt^［19］	67.2	87.3	76.4	55.3
Point-MLPBLS	68.1	88.7	79.1	31.7

倒置残差模块	特征融合模块	BLS多层感知机级联结构	激活函数ReLU	OA/%
				86.3
√				87.2
√	√			87.6
√	√	√		88.4
√	√	√	√	88.7
	√			86.8
		√		87.0
			√	86.5
	√	√		87.5
√		√	√	88.5
	√	√	√	88.1

Point-MLPBLS： point cloud semantic segmentation network based on MLP cascaded broad learning system

基于多层感知机级联宽度学习系统的点云语义分割网络Point-MLPBLS

RichHTML

PDF

Knowledge

Abstract

Cite this article

share this article

Figures/Tables 10

References 19

Related Articles 15

Recommended Articles

Metrics

[1]	Liu WANG, Xuebin CHEN, Yuan GAO, Kaiguang MA, Tong ZHAO. Dual-channel recommendation algorithm based on FFT denoising and reverse path collaborative enhancement [J]. Journal of Computer Applications, 2026, 46(7): 2130-2138.
[2]	Yasong ZHANG, Bihui CONG, Shuang XU. Graph neural network node classification model incorporating clustering coefficients [J]. Journal of Computer Applications, 2026, 46(6): 1855-1862.
[3]	Jinxiao ZHANG, Chenglong LI, Xinyan GAO, Ming ZHANG. 3D human pose estimation model based on temporal-spatial feature pyramid network and multi-hypothesis interaction mechanism [J]. Journal of Computer Applications, 2026, 46(6): 1965-1972.
[4]	Yunping HE, Leichun WANG, Ruirui SONG, Xiangfeng LU, Jinxiang WEI, Xiaomeng LIU. Dual-channel multimodal sentiment analysis model based on contrast invariance and reinforcement specificity [J]. Journal of Computer Applications, 2026, 46(6): 1767-1775.
[5]	Liwan YAO, Hailong LIU, Zhangfan ZENG. Frequency-domain driven and diffusion-based fusion for sonar image enhancement algorithm [J]. Journal of Computer Applications, 2026, 46(6): 1947-1955.
[6]	Songhao ZHU, Zhiyun ZHAO, Mengling WANG. Low-light image enhancement network based on lightweight residual and brightness-aware dynamic feature fusion [J]. Journal of Computer Applications, 2026, 46(6): 1936-1946.
[7]	Chao LYU, Geyao MA. Lightweight human pose estimation network based on redundant feature suppression [J]. Journal of Computer Applications, 2026, 46(6): 1973-1980.
[8]	Jun QIN, Xintao JIAO, Biqing ZENG. Non-intrusive load monitoring method combining BiLSTM-Transformer and Kolmogorov-Arnold network [J]. Journal of Computer Applications, 2026, 46(6): 2026-2033.
[9]	Jing HU, Shikun CHEN, Fang WANG, Rui ZHANG, Yong WANG. Ore image segmentation with linear deformable convolution and dual-domain synergistic dynamic attention [J]. Journal of Computer Applications, 2026, 46(5): 1692-1702.
[10]	Qianfei WANG, Yang LI, Deyu LI, Suge WANG. Dual-channel feature fusion representation method for short-text clustering based on large language model [J]. Journal of Computer Applications, 2026, 46(5): 1441-1449.
[11]	Xuechao LIAO, Rui CHEN. Prediction-evaluation framework for anomaly detection in electric vehicle lithium-ion battery [J]. Journal of Computer Applications, 2026, 46(5): 1614-1623.
[12]	Huaze ZHU, Weihao WANG, Mingyu YOU, Hongjun ZHOU. 3D part assembly method based on line drawing segmentation [J]. Journal of Computer Applications, 2026, 46(5): 1545-1550.
[13]	Ying JING, Ran LI, Zhuo JIANG, Ziyang FU, Jingyi DU, Qi LIU, Jihang LIU. SAM Meibomian gland unified dense segmentation method with introduction of automatic prompt encoder [J]. Journal of Computer Applications, 2026, 46(5): 1667-1676.
[14]	Baoyuan ZHENG, Chaobo HE. Graph convolutional network enhanced by graph diffusion and dual-view feature learning [J]. Journal of Computer Applications, 2026, 46(5): 1370-1377.
[15]	Minqi WU, Yuanhua YANG, Hang LI, Yaqin HU, Zhihao TANG, Teng MEI. Lightweight underwater small object detection based on graph Transformer and RT-DETR [J]. Journal of Computer Applications, 2026, 46(5): 1586-1595.

全连接网络	BLS	MLPBLS	mAcc/%	分割时间/min
√			79.4	23.3
	√		77.6	2.0
		√	79.1	2.6

全连接网络	BLS	MLPBLS	mAcc/%	分割时间/min
√			79.4	23.3
	√		77.6	2.0
		√	79.1	2.6