| 引用本文: | 焦倩,马飞,王志伟,杨岩立,郑莉芳,刘博深.雪道点云时域波形特征分割及SLAM算法[J].哈尔滨工业大学学报,2024,56(8):124.DOI:10.11918/202307071 |
| JIAO Qian,MA Fei,WANG Zhiwei,YANG Yanli,ZHENG Lifang,LIU Boshen.Ski pistes segmentation and SLAM algorithm based on time-domain waveform characteristics of LiDAR point cloud[J].Journal of Harbin Institute of Technology,2024,56(8):124.DOI:10.11918/202307071 |
|
| |
|
|
| 本文已被：浏览 1134次 下载 1394次 |
码上扫一扫！ |
|
|
| 雪道点云时域波形特征分割及SLAM算法 |
|
焦倩1,马飞1,2,王志伟1,杨岩立3,郑莉芳1,刘博深1,2
|
|
(1.北京科技大学 机械工程学院,北京 100083; 2.北京科技大学 顺德创新学院,广东 佛山 528339; 3.河北宣工机械发展有限责任公司,河北 张家口 075105)
|
|
| 摘要: |
| 为提升压雪车在滑雪场行驶与作业的夜间环境感知能力,提出了压雪车车载激光雷达点云时域波形阶跃值检测的雪道分割算法,并构建了适用于高山滑雪场的实时定位与建图算法（SLAM）。首先,对点云反射率分布进行统计并标记飘雪噪声点,利用邻近点线性插值方法对标记的飘雪噪声进行处理,保证点云的时域波形连续性,同时在特定扫描视角内划分栅格并以相邻网格高程值变化实时估计坡度。其次,根据山地滑雪场典型障碍设计了相应的时域波形阶跃值检测判据,筛选雪道-障碍分界点,划分扫描区间并对区间内的波形阶跃值进行包络,实现雪道点云特征分割。然后,根据雪道分割结果对特征进行分类匹配,并利用特征约束方法提高建图速度。最后,在张家口万龙滑雪场的高级和中级雪道进行算法效果验证测试。测试结果表明:所提出的雪道分割算法对单帧点云数据的处理平均耗时为2.36 ms,平均分割准确率可达98.54%,在基于雪道分割算法的改进SLAM方法中可以准确实现雪道-障碍分割,在建图精度方面表现更为优越且大幅缩减了计算耗时。 |
| 关键词: 高山滑雪场 激光雷达 飘雪噪声 时域波形 雪道分割 SLAM |
| DOI：10.11918/202307071 |
| 分类号:TN959.71 |
| 文献标识码:A |
| 基金项目:国家重点研发计划项目(2020YFF0304000) |
|
| Ski pistes segmentation and SLAM algorithm based on time-domain waveform characteristics of LiDAR point cloud |
|
JIAO Qian1,MA Fei1,2,WANG Zhiwei1,YANG Yanli 3,ZHENG Lifang1,LIU Boshen1,2
|
|
(1.School of Mechanical and Engineering, University of Science and Technology Beijing, Beijing 100083, China; 2.Shunde Innovation School, University of Science and Technology Beijing, Foshan 528339, Guangdong, China; 3.Hebei Xuanhua Construction Machinery Co., Ltd., Zhangjiakou 075105, Hebei, China)
|
| Abstract: |
| To enhance the perception ability of snow field environments during nighttime and improve both the maintenance quality and operational efficiency of ski pistes, this paper proposes a novel segmentation algorithm based on time-domain waveform characteristics detection using vehicle-mounted LiDAR for the snow groomer and an improved simultaneous localization and mapping (SLAM) algorithm in ski pistes. First, based on a model of reflectivity distribution of the point cloud, snow noise points are identified and processed using linear interpolation to maintain the continuity of the time-domain waveform. Consequently, the grid is partitioned into specific scan perspectives to accurately estimate the real-time slope based on the elevation variation between neighboring grids. Then, a corresponding criterion for detecting step values in the time-domain waveforms is established to select the boundary points between pistes and obstacles according to the typical obstacles in alpine ski resorts. The segmentation can be achieved by enveloping the step values. Moreover, based on the segmentation results, features are classified and matched using feature constraint methods to improve the mapping speed. Finally, tests were conducted on the pistes of the Wanlong Ski Resort in Zhangjiakou. Experimental results show that the proposed algorithm for segmenting ski pistes achieves an average processing time of 2.36 ms per single-frame point cloud, while achieving an accuracy rate of over 98.54%. Moreover, when integrated into an improved SLAM approach with snow piste segmentation capabilities, it can achieve segmentation accurately while demonstrating superior mapping accuracy performance, as well as significantly reduced computational time. |
| Key words: alpine ski resorts vehicle-mounted LiDAR snow noise time-domain waveform ski pistes segmentation SLAM |
|
|
|
|
