[1]刘起源,路锦正,黄炳森.基于特征融合和损失优化的点云语义分割网络[J].计算机技术与发展,2024,34(05):66-72.[doi:10.20165/j.cnki.ISSN1673-629X.2024.0042]
　LIU Qi-yuan,LU Jin-zheng,HUANG Bing-sen.Point Cloud Semantic Segmentation Network Based on Feature Fusion and Loss Optimization[J].,2024,34(05):66-72.[doi:10.20165/j.cnki.ISSN1673-629X.2024.0042]
点击复制

基于特征融合和损失优化的点云语义分割网络()

《计算机技术与发展》[ISSN:1006-6977/CN:61-1281/TN]

卷:
34
期数:
2024年05期
页码:
66-72
栏目:
媒体计算
出版日期:
2024-05-10

文章信息/Info

Title:
Point Cloud Semantic Segmentation Network Based on Feature Fusion and Loss Optimization
文章编号:
1673-629X(2024)05-0066-07
作者:
刘起源1; 路锦正2; 黄炳森1
1. 西南科技大学 计算机科学与技术学院,四川 绵阳 621010;2. 西南科技大学 信息工程学院,四川 绵阳 621010
Author(s):
LIU Qi-yuan1; LU Jin-zheng2; HUANG Bing-sen1
1. School of Computer Science and Technology,Southwest University of Science and Technology,Mianyang 621010,China;2. School of Information Engineering,Southwest University of Science and Technology,Mianyang 621010,China
关键词:
点云; 语义分割; 多尺度特征融合; 损失优化; 神经网络优化
Keywords:
point cloud; semantic segmentation; multi-level feature fusion; loss optimization; neural network optimization
分类号:
TP391.4
DOI:
10.20165/j.cnki.ISSN1673-629X.2024.0042
摘要:
针对目前大多数方法仅利用单尺度特征而忽视了具有不同感受野的多尺度特征信息、无法有效处理点云数据集中类别权重不平衡的问题,提出一种基于全阶段特征融合(FSFF)和平衡损失(BL)的分割网络(FFBL-Net)。 首先,FSFF模块通过将不同编码阶段的可学习特征与当前阶段特征进行融合,促进了浅层和深层语义信息互补;融合后的特征被传递到编码融合模块(EFM)和解码融合模块(DFM),实现了特征的跨阶段融合。 此外,为了解决数据集中类别分布不平衡的问题,引入 BL 损失调整类别间的梯度差异。 实验结果表明,FFBL-Net 在主流的大规模点云数据集 S3DIS 上,平均交并比达到了 69. 7% ,总体准确率达到了 89. 9% 。 与 PointNet++相比,FFBL-Net 分别提升了 12. 4% 和 6. 1% 。
Abstract:
Aiming at the problem that most of the current methods only use single - scale features but ignore the multi - scale feature information with different receptive fields and cannot effectively deal with unbalanced category weights in point cloud datasets, a segmentation network (FFBL-Net) based on full-stage feature fusion (FSFF) and balanced loss (BL) is proposed. First,FSFF module promotes the complementation of shallow and deep semantic information by integrating learnable features of different coding stages with features of the current stage. The fused features are transferred to the encoding fusion module (EFM) and decoding fusion module(DFM),which realizes the cross - stage fusion of features. In addition, to solve the problem of unbalanced class distribution in the dataset,BL loss is introduced to adjust the gradient difference between categories. The experimental results show that the FFBL-Net onthe large-scale point cloud dataset S3DIS has reached 69. 7% in terms of mean intersection over union (mIoU) and 89. 9% in overall accuracy (OA),which is 12. 4% and 6. 1% higher than that of the original PointNet++ respectively.

相似文献/References:

[1]赵越超 李忠科 王勇 吕培军.基于OpenGL的三维牙颌模型可视化研究[J].计算机技术与发展,2008,(01):119.
　ZHAO Yue-chao,LI Zhong-ke,WANG Yong,et al.Research of 3D Dental Visualization Based on OpenGL[J].,2008,(05):119.
[2]张琴 蔡勇.支持向量学习机在点云去噪中的应用[J].计算机技术与发展,2011,(06):85.
　ZHANG Qin,CAI Yong.Application of Support Vector Machine in Point Clouds Denoising[J].,2011,(05):85.
[3]周春艳 李勇 邹峥嵘.三维点云ICP算法改进研究[J].计算机技术与发展,2011,(08):75.
　ZHOU Chun-yan,LI Yong,ZOU Zheng-rong.Three-Dimensional Cloud ICP Algorithm Improvement[J].,2011,(05):75.
[4]谢晓燕,吴锦桥. 一种全自动三维点云配准及比例约束方法[J].计算机技术与发展,2015,25(03):63.
　XIE Xiao-yan,WU Jin-qiao. An Automatic Method of 3 D Point Cloud Registration and Dimension Adjustment[J].,2015,25(05):63.
[5]廖中平,白慧鹏,陈 立.基于双边滤波改进的点云平滑算法[J].计算机技术与发展,2019,29(11):42.[doi:10. 3969 / j. issn. 1673-629X. 2019. 11. 009]
　LIAO Zhong-ping,BAI Hui-peng,CHEN Li.Improved Denoising of Point-sampled Model Based on Bilateral Filtering[J].,2019,29(05):42.[doi:10. 3969 / j. issn. 1673-629X. 2019. 11. 009]
[6]刘保安.ICP 算法在文物三维重建中的应用[J].计算机技术与发展,2020,30(01):217.[doi:10. 3969 / j. issn. 1673-629X. 2020. 01. 039]
　LIU Bao-an.Application of ICP Algorithm in Three-dimensional Reconstruction of Cultural Relics[J].,2020,30(05):217.[doi:10. 3969 / j. issn. 1673-629X. 2020. 01. 039]
[7]钟 诚,周浩杰,韦海亮.一种基于注意力机制的三维点云物体识别方法[J].计算机技术与发展,2020,30(04):41.[doi:10. 3969 / j. issn. 1673-629X. 2020. 04. 008]
　ZHONG Cheng,ZHOU Hao-jie,WEI Hai-liang.A 3D Point Cloud Object Recognition Method Based on Attention Mechanism[J].,2020,30(05):41.[doi:10. 3969 / j. issn. 1673-629X. 2020. 04. 008]
[8]张泽宇,郭 斌,张太红*.基于 DCNN 的马匹图像分割算法研究[J].计算机技术与发展,2020,30(10):210.[doi:10. 3969 / j. issn. 1673-629X. 2020. 10. 037]
　ZHANG Ze-yu,GUO Bin,ZHANG Tai-hong.Research on Horse Image Segmentation Algorithm Based on DCNN[J].,2020,30(05):210.[doi:10. 3969 / j. issn. 1673-629X. 2020. 10. 037]
[9]龚 安,赵 莉,姚鑫杰.基于改进的 U-Net 网络模型的气胸分割算法[J].计算机技术与发展,2021,31(10):173.[doi:10. 3969 / j. issn. 1673-629X. 2021. 10. 029]
　GONG An,ZHAO Li,YAO Xin-jie.Pneumothorax Segmentation Algorithm Based on Improved U-Net Network Model[J].,2021,31(05):173.[doi:10. 3969 / j. issn. 1673-629X. 2021. 10. 029]
[10]宋 宇,王小瑀,梁 超,等.基于多级特征图联合上采样的实时语义分割[J].计算机技术与发展,2022,32(02):82.[doi:10. 3969 / j. issn. 1673-629X. 2022. 02. 013]
　SONG Yu,WANG Xiao-yu,LIANG Chao,et al.Real-time Semantic Segmentation Based on Multi-scale Feature Map Joint Pyramid Upsamping[J].,2022,32(05):82.[doi:10. 3969 / j. issn. 1673-629X. 2022. 02. 013]

更新日期/Last Update: 2024-05-10