Giter Club home page Giter Club logo

lidar_denoise's Introduction

Lidar Denoise 轻量级激光雷达点云降噪网络

For more details on the research behind this project, please refer to the following paper:

  • Title: LAPRNet: Lightweight Airborne Particle Removal Network for LiDAR Point Clouds
  • Link: Springer Link

Please cite this paper if you use our project in your research.

简介

本项目提出了一种新的轻量级网络,用于过滤恶劣天气条件下激光雷达点云中的噪点。该网络采用宽度多级残差模块(WMLR)架构,集成了宽激活、多级残差连接和shuffle attention机制,以实现高效的点云预处理。此外,本文还介绍了增强型激光雷达数据表示,以提升系统性能。

主要内容

车辆的自动驾驶和移动机器人依赖激光雷达传感器进行室外环境感知。但在雾、雨、雪等恶劣天气条件下,空中颗粒物会引入不必要的测量点,导致漏检和误报。传统的雷达点云降噪方法和基于深度学习的方法都存在一定的局限性。为此,本文提出了一种轻量级网络,旨在改善基于激光雷达的感知系统在恶劣天气中的性能。

网络架构

本文的网络架构包括三个宽度多级残差模块(WMLR),具有以下特点:

  • 宽激活,提供更丰富的特征表示。
  • 多级残差连接,增强了网络的学习能力。
  • Shuffle attention机制,提高了特征的利用效率。

数据表示

提出了增强型激光雷达数据表示,结合了点云空间分布、标准强度和距离输入,提高了性能。

模型

提出了两种模型:$LAPRNet_2$ 和$LAPRNet_3$,它们遵循相同的网络架构但具有不同的输入表示。它们在受控室内环境和自然天气环境下进行训练和测试。

性能

通过WADS和Chamber数据集的实验表明,提出的模型性能优于现有的深度学习和传统滤波方法。同时,在计算资源有限的边缘设备上也能够实现最佳的性能和计算成本平衡。

表格1: Comparison of runtime and model complexity

Model Params (Mio) GFLOPs Runtime (ms) Size (M)
DROR $4e^{-6}$ - 100.00 $4e^{-6}$
LilaNet 9.31 447.67 5.9417 37.34
WeatherNet 1.5313 18.2267 2.22 6.044
PolarNet 13.6091 87.9441 22.11 54.523
$LAPRNet_2$ 0.3884 4.9119 1.64 1.568
$LAPRNet_3$ 0.3884 4.9919 10.32 1.568

*Comparison of runtime and model complexity. The best method in bold and the second best in italic.

表格2: Comparison on the Chamber Dataset

Model IoU Clear IoU Fog IoU Rain mIoU Precision Recall
DROR 88.13 6.94 7.37 34.15 - -
LiLanet 82.72 79.57 88.16 83.48 - -
WeatherNet 91.65 86.40 89.29 89.11 89.87 92.23
PolarNet 99.08 94.75 91.35 95.06 96.33 97.00
$LAPRNet_2$ 97.17 91.58 94.80 94.51 95.46 97.99
$LAPRNet_3$ 97.83 93.93 95.13 95.63 93.04 98.13

*Comparison on the Chamber Dataset. The best method in bold and the second best in italic.

表格3: Comparison on the WADS Dataset

Model IoU Clear IoU Snow mIoU Precision Recall
DROR - 67.26 - 71.51 91.89
DSOR - 63.18 - 65.07 95.60
WeatherNet 92.29 81.39 86.84 94.88 85.13
PolarNet 98.72 90.34 94.53 98.74 91.41
$LAPRNet_2$ 94.73 87.31 91.02 95.42 91.13
$LAPRNet_3$ 95.65 89.53 92.59 95.95 93.04

*Comparison on the WADS Dataset. The best method in bold and the second best in italic.

图像1: 网络架构

网络架构

图像2: 性能对比

性能对比

lidar_denoise's People

Contributors

qizhi697 avatar

Watchers

 avatar

Recommend Projects

  • React photo React

    A declarative, efficient, and flexible JavaScript library for building user interfaces.

  • Vue.js photo Vue.js

    🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.

  • Typescript photo Typescript

    TypeScript is a superset of JavaScript that compiles to clean JavaScript output.

  • TensorFlow photo TensorFlow

    An Open Source Machine Learning Framework for Everyone

  • Django photo Django

    The Web framework for perfectionists with deadlines.

  • D3 photo D3

    Bring data to life with SVG, Canvas and HTML. 📊📈🎉

Recommend Topics

  • javascript

    JavaScript (JS) is a lightweight interpreted programming language with first-class functions.

  • web

    Some thing interesting about web. New door for the world.

  • server

    A server is a program made to process requests and deliver data to clients.

  • Machine learning

    Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.

  • Game

    Some thing interesting about game, make everyone happy.

Recommend Org

  • Facebook photo Facebook

    We are working to build community through open source technology. NB: members must have two-factor auth.

  • Microsoft photo Microsoft

    Open source projects and samples from Microsoft.

  • Google photo Google

    Google ❤️ Open Source for everyone.

  • D3 photo D3

    Data-Driven Documents codes.