| 摘 要: 为解决多线束激光雷达原始数据包含大量噪声以及数据量大较难实时处理的问题，本文基于Qt框架和OpenGL图形库设计开发了一套点云实时滤波与可视化系统。软件系统基于生产者——消费者多线程模型实现数据接收和数据处理的并行操作，在点云滤波模块通过多核CPU并行计算实现每秒对百万级数据点的滤波，基于OpenGL开发三维点云可视化模块，同时支持设备交互和信息显示。实验表示，相较于传统激光雷达500米以内的探测距离，单光子激光雷达系统的有效探测距离可达到2500米，多线程滤波相较于单线程滤波有10倍左右性能提升，渲染百万级三维点云仅需0.05秒，在三维测绘、海事应用等领域具有较高的应用价值。 |
| 关键词: 激光雷达 并行计算 点云滤波 三维点云可视化 OpenGL |
|
中图分类号:
文献标识码:
|
|
| Design of Point Cloud Filtering and Visualization System for Single-photon Multi-beam LiDAR |
|
zhaowenhao1, pengziqiang2, shexiaokai2
|
1.Shanghai University of Engineering Science;2.Shanghai Institute of Technical Physics, Chinese Academy of Sciences
|
| Abstract: To address the heavy background noise and the large data volume that hinder real-time processing of raw data from multi-line LiDAR, we designed and implemented a real-time point-cloud filtering and visualization system built on the Qt framework and the OpenGL graphics library. The software adopts a producer–consumer multithreaded model to parallelize data reception and processing. In the point-cloud filtering module, multicore CPU parallelization supports a throughput of about one million points per second. A three-dimensional point-cloud visualization module implemented with OpenGL provides sensor interaction and on-screen information display.Experimental results show that, compared with conventional LiDAR whose effective detection range is within five hundred meters, the single-photon LiDAR system reaches two thousand five hundred meters. Multithreaded filtering is about ten times faster than single-threaded filtering, and rendering a million-point three-dimensional cloud takes about fifty milliseconds per frame. These capabilities indicate strong applicability to three-dimensional mapping and maritime scenarios. |
| Keywords: LiDAR Parallel computing Point-cloud filtering 3D point-cloud visualization OpenGL |
