基于邻域特征编码优化的液压支架激光点云分割算法

王俊甫; 薛晓杰; 杨艺

doi:10.13272/j.issn.1671-251x.2024040052

基于邻域特征编码优化的液压支架激光点云分割算法

doi: 10.13272/j.issn.1671-251x.2024040052

王俊甫^1,,
薛晓杰^1, ,,
杨艺^{2, 3}

1.
郑州恒达智控科技股份有限公司，河南郑州　450000
2.
河南理工大学电气工程与自动化学院，河南焦作　454003
3.
河南理工大学河南省煤矿装备智能检测与控制重点实验室，河南焦作　454003

基金项目: 河南省科技攻关项目（232102210040）。

详细信息

作者简介:
王俊甫（1982—），男，河南濮阳人，工程师，硕士，现从事智能开采装备研发方面的工作，E-mail：wangjunfu@hdzk.com

通讯作者:
薛晓杰（1999—），男，河南郑州人，硕士，现从事人工智能和三维点云语义分割方面的工作，E-mail:Jeremy648@163.com。

中图分类号: TD355
计量
- 文章访问数: 83
- HTML全文浏览量: 30
- PDF下载量: 15
- 被引次数: 0
出版历程
- 收稿日期: 2024-04-17
- 修回日期: 2024-07-28
- 网络出版日期: 2024-08-01

Laser point cloud segmentation algorithm for hydraulic support based on neighborhood feature encoding and optimization

WANG Junfu^1
,,
XUE Xiaojie^{1
, ,},
YANG Yi^{2, 3}

1.
Zhengzhou Hengda Intelligent Control Technology Company Limited, Zhengzhou 450000, China
2.
School of Electrical Engineering and Automation, Henan Polytechnic University, Jiaozuo 454003, China
3.
Henan Key Laboratory of Intelligent Detection and Control of Coal Mine Equipment, Henan Polytechnic University, Jiaozuo 454003, China

摘要

摘要: 受井下煤尘和易被遮挡的影响，液压支架激光点云数据容易出现残缺。现有点云分割算法难以获取细粒度的点云特征，无法得到完整的点云结构信息，且易在邻域内引入语义信息不相似的点，导致液压支架激光点云分割精度低。针对上述问题，提出了一种基于邻域特征编码优化的液压支架激光点云分割算法。引入了由邻域特征编码模块、邻域特征优化模块和混合池化模块组成的局部邻域特征聚合模块：邻域特征编码模块在传统三维坐标编码的基础上加入极坐标编码和质心偏移来表征局部点云空间结构，提升对残缺点云的特征提取能力；邻域特征优化模块通过特征距离判断并丢弃冗余特征，来优化邻域空间内的特征表达，从而更有效地学习点云局部细粒度特征，增强点云局部上下文信息；混合池化模块结合注意力池化和最大池化，通过聚合邻域内的显著特征和重要特征来获取具有丰富信息的单点特征，减少信息丢失。构建了由2组局部邻域特征聚合模块和残差连接组成的邻域扩张模块，以捕获特征间的长距离依赖关系，扩大单个点的局部感受野，并聚合更多有效特征。实验结果表明，该算法在液压支架激光点云分割数据集上的平均交并比为93.26%，平均准确率为96.42%，可有效区分液压支架不同的几何结构，实现液压支架各部件的准确分割。
- 液压支架 /
- 激光点云 /
- 点云分割 /
- 邻域特征编码 /
- 邻域特征优化
Abstract: Due to the influence of underground coal dust and easy obstruction, the laser point cloud data of hydraulic supports is prone to be incomplete. The existing point cloud segmentation algorithms are difficult to obtain fine-grained point cloud features, unable to obtain complete structural information of the point cloud. The algorithms are prone to introducing semantically dissimilar points in the neighborhood, resulting in low precision of laser point cloud segmentation for hydraulic supports. In order to solve the above problems, a laser point cloud segmentation algorithm for hydraulic supports based on neighborhood feature encoding and optimization is proposed. The method introduces a local neighborhood feature aggregation module consisting of neighborhood feature encoding module, neighborhood feature optimization module, and hybrid pooling module. The neighborhood feature encoding module adds polar coordinate encoding and centroid offset to represent the spatial structure of local point clouds on the basis of traditional 3D coordinate encoding, improving the feature extraction capability for incomplete point clouds. The neighborhood feature optimization module optimizes the feature expression in the neighborhood space by judging the feature distance and discarding redundant features, thereby more effectively learning the local fine-grained features of the point cloud and enhancing the local contextual information of the point cloud. The hybrid pooling module combines attention pooling and max pooling to obtain single point features with rich information by aggregating salient and important features within the neighborhood, reducing information loss. A neighborhood expansion module consisting of two sets of local neighborhood feature aggregation modules and residual connections is constructed to capture long-range dependencies between features, expand the local receptive field of individual points, and aggregate more effective features. The experimental results show that the algorithm has an mean intersection over union of 93.26% and an average accuracy of 96.42% on the laser point cloud segmentation dataset of hydraulic supports. It can effectively distinguish different geometric structures of hydraulic supports and achieve accurate segmentation of various components of hydraulic supports.
- hydraulic support /
- laser point cloud /
- point cloud segmentation /
- neighborhood feature encoding /
- neighborhood feature optimization

HTML全文

图 1 基于邻域特征编码优化的液压支架激光点云分割算法框架

Figure 1. Architecture of laser point cloud segmentation algorithm for hydraulic support based on neighborhood feature encoding and optimization

下载: 全尺寸图片幻灯片

图 2 邻域扩张模块结构

Figure 2. Structure of neighborhood expanding module

下载: 全尺寸图片幻灯片

图 3 邻域特征编码模块结构

Figure 3. Structure of neighborhood feature encoding module

下载: 全尺寸图片幻灯片

图 4 邻域点云极坐标转换

Figure 4. Polar coordinate transformation of neighborhood point cloud

下载: 全尺寸图片幻灯片

图 5 质心偏移

Figure 5. Centroid offset

下载: 全尺寸图片幻灯片

图 6 邻域特征优化模块结构

Figure 6. Structure of neighborhood feature optimization module

下载: 全尺寸图片幻灯片

图 7 混合池化模块结构

Figure 7. Structure of mixed pooling module

下载: 全尺寸图片幻灯片

图 8 数据增强方式

Figure 8. Data enhancement mode

下载: 全尺寸图片幻灯片

图 9 液压支架结构

Figure 9. Hydraulic support structure

下载: 全尺寸图片幻灯片

图 10 不同算法整体分割结果可视化对比

Figure 10. Visual comparison of overall segmentation results by different algorithms

下载: 全尺寸图片幻灯片

图 11 不同算法部件分割结果可视化对比

Figure 11. Visual comparison of component segmentation results by different algorithms

下载: 全尺寸图片幻灯片

表 1 不同算法评价指标对比

Table 1. Comparison of evaluation indexes of different algorithms %

算法	mAcc	OA	mIoU
RandLA−Net	95.93	96.31	92.51
本文算法	96.42	96.83	93.26

下载: 导出CSV

表 2 不同算法在各类别上的IoU对比

Table 2. Intersection over union（IoU） comparison of different algorithms in various categories %

算法	IoU
算法	掩护梁	立柱	顶梁	其他
RandLA−Net	92.99	88.78	93.23	95
本文算法	93.96	89.28	93.79	96

下载: 导出CSV

表 3 消融实验结果

Table 3. Results of ablation experiments

算法	邻域特征编码模块	邻域特征优化模块	混合池化模块	邻域扩张模块	mIoU/%
1	×	×	×	×	91.97
2	√	×	×	×	92.50
3	√	√	×	×	92.76
4	√	√	√	×	92.90
5	√	√	√	√	93.26

下载: 导出CSV

参考文献(21)

[1]	王国法,刘峰,孟祥军,等. 煤矿智能化(初级阶段)研究与实践[J]. 煤炭科学技术,2019,47(8):1-36. WANG Guofa,LIU Feng,MENG Xiangjun,et al. Research and practice of coal mine intellectualization (primary stage)[J]. Coal Science and Technology,2019,47(8):1-36.
[2]	GUO Jun,HUANG Wenbo,FENG Guorui,et al. Stability analysis of longwall top-coal caving face in extra-thick coal seams based on an innovative numerical hydraulic support model[J]. International Journal of Mining Science and Technology,2024,34(4):491-505. doi: 10.1016/j.ijmst.2024.04.011
[3]	高有进,杨艺,常亚军,等. 综采工作面智能化关键技术现状与展望[J]. 煤炭科学技术,2021,49(8):1-22. GAO Youjin,YANG Yi,CHANG Yajun,et al. Status and prospect of key technologies of intelligentization of fully-mechanized coal mining face[J]. Coal Science and Technology,2021,49(8):1-22.
[4]	王国法,庞义辉,许永祥,等. 厚煤层智能绿色高效开采技术与装备研发进展[J]. 采矿与安全工程学报,2023,40(5):882-893. WANG Guofa,PANG Yihui,XU Yongxiang,et al. Development of intelligent green and efficient mining technology and equipment for thick coal seam[J]. Journal of Mining & Safety Engineering,2023,40(5):882-893.
[5]	王国法,庞义辉,任怀伟,等. 智慧矿山系统工程及关键技术研究与实践[J]. 煤炭学报,2024,49(1):181-202. doi: 10.13225/j.cnki.jccs.2023.1355 WANG Guofa,PANG Yihui,REN Huaiwei,et al. System engineering and key technologies research and practice of smart mine[J]. Journal of China Coal Society,2024,49(1):181-202. doi: 10.13225/j.cnki.jccs.2023.1355
[6]	李建,任怀伟,巩师鑫. 综采工作面液压支架状态感知与分析技术研究[J]. 工矿自动化,2023,49(10):1-7,103. LI Jian,REN Huaiwei,GONG Shixin. Research on state perception and analysis technology of hydraulic support in fully mechanized working face[J]. Journal of Mine Automation,2023,49(10):1-7,103.
[7]	王国法,杜毅博. 智慧煤矿与智能化开采技术的发展方向[J]. 煤炭科学技术,2019,47(1):1-10. WANG Guofa,DU Yibo. Development direction of intelligent coal mine and intelligent mining technology[J]. Coal Science and Technology,2019,47(1):1-10.
[8]	XI Xiaohuan,WAN Yiping,WANG Cheng. Building boundaries extraction from points cloud using an image edge detection method[C]. IEEE International Geoscience and Remote Sensing Symposium,Beijing,2016:1270-1273.
[9]	SCHNABEL R,WAHL R,KLEIN R. Efficient RANSAC for point-cloud shape detection[J]. Computer Graphics Forum,2007,26(2):214-226. doi: 10.1111/j.1467-8659.2007.01016.x
[10]	ZHOU Dingfu,FANG Jin,SONG Xibin,et al. Joint 3D instance segmentation and object detection for autonomous driving[C]. IEEE/CVF Conference on Computer Vision and Pattern Recognition,Seattle,2020:1836-1846.
[11]	XU Yongyang,TANG Wei,ZENG Ziyin,et al. NeiEA-NET:semantic segmentation of large-scale point cloud scene via neighbor enhancement and aggregation[J]. International Journal of Applied Earth Observation and Geoinformation,2023,119. DOI: 10.1016/j.jag.2023.103285.
[12]	MATURANA D,SCHERER S. VoxNet:a 3D convolutional neural network for real-time object recognition[C]. IEEE/RSJ International Conference on Intelligent Robots and Systems,Hamburg,2015:922-928.
[13]	LE T,DUAN Ye. PointGrid:a deep network for 3D shape understanding[C]. IEEE/CVF Conference on Computer Vision and Pattern Recognition,Salt Lake City,2018:9204-9214.
[14]	SU Hang,MAJI S,KALOGERAKIS E,et al. Multi-view convolutional neural networks for 3D shape recognition[C]. IEEE International Conference on Computer Vision,Santiago,2015:945-953.
[15]	MILIOTO A,VIZZO I,BEHLEY J,et al. RangeNet:fast and accurate LiDAR semantic segmentation[C]. IEEE/RSJ International Conference on Intelligent Robots and Systems,Macau,2019:4213-4220.
[16]	CHARLES R Q,HAO Su,MO Kaichun,et al. PointNet:deep learning on point sets for 3D classification and segmentation[C]. IEEE Conference on Computer Vision and Pattern Recognition,Honolulu,2017:77-85.
[17]	QI C R,YI Li,SU Hao,et al. PointNet++:deep hierarchical feature learning on point sets in a metric space[C]. The 31st International Conference on Neural Information Processing Systems,Long Beach,2017:5105-5114.
[18]	HU Qingyong,YANG Bo,XIE Linhai,et al. RandLA-Net:efficient semantic segmentation of large-scale point clouds[C]. IEEE/CVF Conference on Computer Vision and Pattern Recognition,Seattle,2020:11105-11114.
[19]	FAN Siqi,DONG Qiulei,ZHU Fenghua,et al. SCF-Net:learning spatial contextual features for large-scale point cloud segmentation[C]. IEEE/CVF Conference on Computer Vision and Pattern Recognition,Nashville,2021:14499-14508.
[20]	ZENG Ziyin,XU Yongyang,XIE Zhong,et al. LEARD-Net:semantic segmentation for large-scale point cloud scene[J]. International Journal of Applied Earth Observation and Geoinformation,2022,112. DOI: 10.1016/j.jag.2022.102953.
[21]	QIAN Wei,XING Weiwei,FEI Shumin. H∞ state estimation for neural networks with general activation function and mixed time-varying delays[J]. IEEE Transactions on Neural Networks and Learning Systems,2021,32(9):3909-3918. doi: 10.1109/TNNLS.2020.3016120