基于时空连续补偿的矿山可通行区域识别方法

代博; 王亚飞; 李若尧; 李泽星; 章翼辰; 张睿韬

doi:10.13272/j.issn.1671-251x.2024050067

基于时空连续补偿的矿山可通行区域识别方法

doi: 10.13272/j.issn.1671-251x.2024050067

1.
上海交通大学机械与动力工程学院，上海　200240
2.
上海交通大学汽车动力与智能控制国家工程研究中心，上海　200240

基金项目: 国家自然科学基金项目（52372417, 52072243）。

详细信息

作者简介:
代博（2000—），男，辽宁鞍山人，硕士研究生，主要研究方向为非结构化道路可通行区域识别，E-mail：daibosjtu@163.com

中图分类号: TD67
计量
- 文章访问数: 73
- HTML全文浏览量: 24
- PDF下载量: 6
- 被引次数: 0
出版历程
- 收稿日期: 2024-05-21
- 修回日期: 2024-10-08
- 网络出版日期: 2024-08-02

Method for identifying passable areas in mines based on spatiotemporal continuous compensation

1.
School of Mechanical Engineering, Shanghai Jiao Tong University, Shanghai 200240, China
2.
National Engineering Research Center of Automotive Power and Control, Shanghai Jiao Tong University,Shanghai 200240, China

摘要

摘要: 可通行区域识别是矿山无人驾驶技术中的重要环节。露天矿山道路场景具有道路边界模糊不清及路面平坦度不一等特征，使用传统同心圆地面分割模型进行矿山道路平面拟合时容易出现可通行区域与车辆不连通及帧间可通行区域识别结果不一致等误分类问题。提出了一种基于时空连续补偿的矿山道路可通行区域识别方法。首先，基于同心圆模型对矿山道路建模，并利用主成分分析方法进行多平面拟合，获取初始可通行区域分割结果；然后，基于空间连通性，分别利用区域生长方法和基于密度的噪声应用空间聚类方法对初始可通行区域进行区域连通性滤波及点连通性滤波，得到符合空间连通性的可通行区域；最后，基于时间区域一致性对不同点云帧中可通行性不一致的不稳定区域进行滤除，先根据正态分布变换方法构建栅格地图，再利用时间稳定权重判断栅格稳定性，最终通过区域栅格投影实现不稳定区域的滤除。矿山场景中测试结果表明：该方法的准确率为93.44%，较现有主流方法提升2.27%；召回率为99.14%，较现有主流方法提升8.26%。该方法不仅在不连通区域中具有良好的空间连通性，还在崎岖区域内具有良好的时序稳定性。
- 露天矿山 /
- 无人驾驶矿卡 /
- 可通行区域识别 /
- 区域连通性滤波 /
- 时空连续补偿 /
- 点云分割
Abstract: Identifying passable areas is a crucial aspect of autonomous driving technology in mining. Open-pit mining road scenes are characterized by unclear road boundaries and varying surface flatness. When using traditional concentric circle ground segmentation models for fitting mining road planes, misclassification issues often arise, such as disconnection between passable areas and vehicles, and inconsistencies in passable area recognition results across frames. This paper proposed a method for identifying passable areas in mining roads based on spatiotemporal continuous compensation. First, the mining road was modeled using a concentric circle model, and principal component analysis was applied for multi-plane fitting to obtain the initial segmentation results of passable areas. Next, based on spatial connectivity, regional connectivity filtering and point connectivity filtering were performed on the initial passable areas using the region-growing algorithm and density-based spatial clustering of applications with noise algorithm, respectively, to obtain passable areas that meet spatial connectivity criteria. Finally, to eliminate unstable regions with inconsistent passability across different point cloud frames, a grid map was constructed based on a normal distribution transformation algorithm, and temporal stability weights were used to assess grid stability, ultimately filtering out unstable regions through regional grid projection. Test results in mining indicated that the proposed method for identifying passable areas achieved an accuracy of 93.44%, representing a 2.27% improvement over existing mainstream algorithms; the recall rate was 99.14%, reflecting an 8.26% enhancement compared to current mainstream algorithms. The proposed method not only exhibits good spatial connectivity in disconnected areas but also demonstrates strong temporal stability in rugged regions.
- open-pit mining /
- autonomous mining trucks /
- passable area identification /
- regional connectivity filtering /
- spatiotemporal continuous compensation /
- point cloud segmentation

HTML全文

图 1 基于时空连续补偿的矿山可通行区域识别方法框架

Figure 1. Method framework of mine passable area identification based on spatiotemporal continuous compensation

下载: 全尺寸图片幻灯片

图 2 无人驾驶矿卡测试平台

Figure 2. Autonomous mining truck platform

下载: 全尺寸图片幻灯片

图 3 铲装区数据采集

Figure 3. Data acquisition of shovel-loading area

下载: 全尺寸图片幻灯片

图 4 点云真值标注示例−俯视

Figure 4. Point cloud annotation example-top view

下载: 全尺寸图片幻灯片

图 5 点云真值标注示例−正视

Figure 5. Point cloud annotation example-front view

下载: 全尺寸图片幻灯片

图 6 可通行区域识别方法效果对比（连通性）

Figure 6. Connectivity of passable area identification algorithms effect comparison

下载: 全尺寸图片幻灯片

图 7 可通行区域识别方法效果对比（稳定性）

Figure 7. Stability effect comparison of passable area identification algorithms

下载: 全尺寸图片幻灯片

表 1 可通行区域识别方法量化对比

Table 1. Quantitative comparison of passable area identification algorithms

方法精度/% 召回率/% 耗时/ms

RANSAC 62.55 74.24 18

Patchwork++ 91.17 90.88 27

本文方法 93.44 99.14 49

下载: 导出CSV

参考文献(23)

[1]	李鑫,余心芷,邱倩茹,等. 无人驾驶矿卡线扫描传感器布设高度优化[J]. 金属矿山,2019,(12):163-167. LI Xin,YU Xinzhi,QIU Qianru,et al. Height layout optimization of line scan sensors installed on unmanned mining dump truck[J]. Metal Mine,2019,(12):163-167.
[2]	ZHOU Zhisong,WANG Yafei,ZHOU Guofeng,et al. A twisted Gaussian risk model considering target vehicle longitudinal-lateral motion states for host vehicle trajectory planning[J]. IEEE Transactions on Intelligent Transportation Systems,2023,24(12):13685-13697. doi: 10.1109/TITS.2023.3298110
[3]	LIU Xulei,WANG Yafei,JIANG Kun,et al. Interactive trajectory prediction using a driving risk map-integrated deep learning method for surrounding vehicles on highways[J]. IEEE Transactions on Intelligent Transportation Systems,2022,23(10):19076-19087. doi: 10.1109/TITS.2022.3160630
[4]	张庚,杨超,王伟达,等. 基于激光雷达的自动驾驶同步定位与建图方法综述[J]. 汽车工程学报,2024,14(1):1-13. ZHANG Geng,YANG Chao,WANG Weida,et al. A review of LiDAR-based simultaneous localization and mapping methods for autonomous driving[J]. Chinese Journal of Automotive Engineering,2024,14(1):1-13.
[5]	DOUILLARD B,UNDERWOOD J,KUNTZ N,et al. On the segmentation of 3D LIDAR point clouds[C]. IEEE International Conference on Robotics and Automation,Shanghai,2011. DOI: 10.1109/ICRA.2011.5979818.
[6]	CHEN Tongtong,DAI Bin,WANG Ruili,et al. Gaussian-process-based real-time ground segmentation for autonomous land vehicles[J]. Journal of Intelligent & Robotic Systems,2014,76(3):563-582.
[7]	TSE R,AHMED N,CAMPBELL M. Unified mixture-model based terrain estimation with Markov Random Fields[C]. IEEE International Conference on Multisensor Fusion and Integration for Intelligent Systems,Hamburg,2012. DOI: 10.1109/MFI.2012.6343027.
[8]	RUMMELHARD L,PAIGWAR A,NEGRE A,et al. Ground estimation and point cloud segmentation using SpatioTemporal Conditional Random Field[C]. IEEE Intelligent Vehicles Symposium ,Los Angeles,2017. DOI: 10.1109/IVS.2017.7995861.
[9]	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. DOI: 10.1109/IROS40897.2019.8967762.
[10]	THRUN S,MONTEMERLO M,DAHLKAMP H,et al. Stanley:the robot that won the DARPA grand challenge[J]. Journal of Field Robotics,2006,23(9):661-692. doi: 10.1002/rob.20147
[11]	ASVADI A,PEIXOTO P,NUNES U. Detection and tracking of moving objects using 2.5 D motion grids[C]. IEEE 18th International Conference on Intelligent Transportation Systems,Gran Canaria,2015. DOI: 10.1109/ITSC.2015.133.
[12]	HIMMELSBACH M,HUNDELSHAUSEN F V,WUENSCHE H J. Fast segmentation of 3D point clouds for ground vehicles[C]. IEEE Intelligent Vehicles Symposium,La Jolla,2010. DOI: 10.1109/IVS.2010.5548059.
[13]	STEINHAUSER D,RUEPP O,BURSCHKA D. Motion segmentation and scene classification from 3D LIDAR data[C]. IEEE Intelligent Vehicles Symposium,Eindhoven,2008. DOI: 10.1109/IVS.2008.4621281.
[14]	ZERMAS D,IZZAT I,PAPANIKOLOPOULOS N. Fast segmentation of 3D point clouds:a paradigm on LiDAR data for autonomous vehicle applications[C]. IEEE International Conference on Robotics and Automation,Singapore,2017. DOI: 10.1109/ICRA.2017.7989591.
[15]	NARKSRI P,TAKEUCHI E,NINOMIYA Y,et al. A slope-robust cascaded ground segmentation in 3D point cloud for autonomous vehicles[C]. 21st International Conference on Intelligent Transportation Systems,Maui,2018. DOI: 10.1109/ITSC.2018.8569534.
[16]	CHENG Jie,HE Dong,LEE C. A simple ground segmentation method for LiDAR 3D point clouds[C]. 2nd International Conference on Advances in Computer Technology,Information Science and Communications,Suzhou,2020. DOI: 10.1109/CTISC49998.2020.00034.
[17]	LIM H,OH M,MYUNG H. Patchwork:concentric zone-based region-wise ground segmentation with ground likelihood estimation using a 3D LiDAR sensor[J]. IEEE Robotics and Automation Letters,2021,6(4):6458-6465. doi: 10.1109/LRA.2021.3093009
[18]	LEE S,LIM H,MYUNG H. Patchwork++:fast and robust ground segmentation solving partial under-segmentation using 3D point cloud[C]. IEEE/RSJ International Conference on Intelligent Robots and Systems,Kyoto,2022. DOI: 10.1109/IR0547612.2022.9981561.
[19]	FENG Chen,TAGUCHI Y,KAMAT V R. Fast plane extraction in organized point clouds using agglomerative hierarchical clustering[C]. IEEE International Conference on Robotics and Automation,Hong Kong,2014. DOI: 10.1109/ICRA.2014.6907776.
[20]	BI Fangming,WANG Weikui,CHEN Long. DBSCAN:Density-based spatial clustering of applications with noise[J]. Journal of Nanjing University(Natural Sciences),2012,48(4):491-498.
[21]	XU Wei,ZHANG Fu. FAST-LIO:a fast,robust LiDAR-inertial odometry package by tightly-coupled iterated Kalman filter[J]. IEEE Robotics and Automation Letters,2021,6(2):3317-3324. doi: 10.1109/LRA.2021.3064227
[22]	SAARINEN J,ANDREASSON H,STOYANOV T,et al. Normal distributions transform occupancy maps:application to large-scale online 3D mapping[C]. IEEE International Conference on Robotics and Automation,Karlsruhe,2013. DOI: 10.1109/ICRA.2013.6630878.
[23]	FISCHLER M A,BOLLES R C. Random sample consensus[J]. Communications of the ACM,1981,24(6):381-395. doi: 10.1145/358669.358692