open3d使用fpfh做点云配准

2024-02-12 00:10
文章标签 使用 配准 做点 fpfh open3d

本文主要是介绍open3d使用fpfh做点云配准,希望对大家解决编程问题提供一定的参考价值,需要的开发者们随着小编来一起学习吧!

目录

  • 写在前面
  • 准备
    • 编译open3d
    • 配准example
  • 编译demo
  • 配准测试
  • 参考

写在前面

1、环境:win10,cmake3.22.0-rc1,环境:win10,cmake3.22.0-rc1,已编译好的open3d 0.13
2、源码使用的是官方example
https://github.com/isl-org/Open3D/blob/master/examples/cpp/RegistrationRANSAC.cpp
3、建议先看一下open3d的demo:
open3d 0.13的c++版本使用demo https://blog.csdn.net/qq_41102371/article/details/12106527
4、所有资源已给出链接

准备

编译open3d

windows10编译open3d 0.13 https://blog.csdn.net/qq_41102371/article/details/121014372

配准example

将官方demo的CMakeLists.txt与配准源码RegistrationRANSAC.cpp
下载下来,放在文件夹open3d_fpfh_registration下
https://github.com/isl-org/open3d-cmake-find-package/blob/master/CMakeLists.txt
https://github.com/isl-org/Open3D/blob/master/examples/cpp/RegistrationRANSAC.cpp
并将CMakeLists.txt里面的Draw全部换成RegistrationRANSAC
CMakeLists.txt

# On Ubuntu 18.04, get the latest CMake from https://apt.kitware.com/.
cmake_minimum_required(VERSION 3.18)project(Open3DCMakeFindPackage LANGUAGES C CXX)# The options need to be the same as Open3D's default
# If Open3D is configured and built with custom options, you'll also need to
# specify the same custom options.
option(STATIC_WINDOWS_RUNTIME "Use static (MT/MTd) Windows runtime" ON)
if(STATIC_WINDOWS_RUNTIME)set(CMAKE_MSVC_RUNTIME_LIBRARY "MultiThreaded$<$<CONFIG:Debug>:Debug>")
else()set(CMAKE_MSVC_RUNTIME_LIBRARY "MultiThreaded$<$<CONFIG:Debug>:Debug>DLL")
endif()# Find installed Open3D, which exports Open3D::Open3D
find_package(Open3D REQUIRED)add_executable(RegistrationRANSAC)
target_sources(RegistrationRANSAC PRIVATE RegistrationRANSAC.cpp)
target_link_libraries(RegistrationRANSAC PRIVATE Open3D::Open3D)# On Windows if BUILD_SHARED_LIBS is enabled, copy .dll files to the executable directory
if(WIN32)get_target_property(open3d_type Open3D::Open3D TYPE)if(open3d_type STREQUAL "SHARED_LIBRARY")message(STATUS "Copying Open3D.dll to ${CMAKE_CURRENT_BINARY_DIR}/$<CONFIG>")add_custom_command(TARGET RegistrationRANSAC POST_BUILDCOMMAND ${CMAKE_COMMAND} -E copy${CMAKE_INSTALL_PREFIX}/bin/Open3D.dll${CMAKE_CURRENT_BINARY_DIR}/$<CONFIG>)endif()
endif()

RegistrationRANSAC.cpp,我在官方代码上做了些修改

// ----------------------------------------------------------------------------
// -                        Open3D: www.open3d.org                            -
// ----------------------------------------------------------------------------
// The MIT License (MIT)
//
// Copyright (c) 2018 www.open3d.org
//
// Permission is hereby granted, free of charge, to any person obtaining a copy
// of this software and associated documentation files (the "Software"), to deal
// in the Software without restriction, including without limitation the rights
// to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
// copies of the Software, and to permit persons to whom the Software is
// furnished to do so, subject to the following conditions:
//
// The above copyright notice and this permission notice shall be included in
// all copies or substantial portions of the Software.
//
// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
// FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
// IN THE SOFTWARE.
// ----------------------------------------------------------------------------#include <Eigen/Dense>
#include <iostream>
#include <memory>#include "open3d/Open3D.h"using namespace open3d;
using namespace std;
std::tuple<std::shared_ptr<geometry::PointCloud>,std::shared_ptr<pipelines::registration::Feature>>
PreprocessPointCloud(const char *file_name, float voxel_size = 2.0) {//从文件读取点云auto pcd = open3d::io::CreatePointCloudFromFile(file_name);//降采样auto pcd_down = pcd->VoxelDownSample(voxel_size);std::cout << "read " << pcd->points_.size() << " points from " << file_name<< std::endl;std::cout << "voxel_size=" << voxel_size << ", after downsample "<< pcd_down->points_.size() << "points left" << endl;//计算法向量pcd_down->EstimateNormals(open3d::geometry::KDTreeSearchParamHybrid(voxel_size * 2, 30));//计算fpfh特征auto pcd_fpfh = pipelines::registration::ComputeFPFHFeature(*pcd_down,open3d::geometry::KDTreeSearchParamHybrid(voxel_size * 5, 100));return std::make_tuple(pcd_down, pcd_fpfh);
}void PrintHelp() {using namespace open3d;PrintOpen3DVersion();// clang-format offutility::LogInfo("Usage:");utility::LogInfo("    > RegistrationRANSAC source_pcd target_pcd  voxel_size [--method=feature_matching] [--mutual_filter] [--visualize]");// clang-format onutility::LogInfo("");
}int main(int argc, char *argv[]) {using namespace open3d;utility::SetVerbosityLevel(utility::VerbosityLevel::Debug);if (argc < 4 ||utility::ProgramOptionExistsAny(argc, argv, {"-h", "--help"})) {PrintHelp();return 1;}bool visualize = false;if (utility::ProgramOptionExists(argc, argv, "--visualize")) {visualize = true;}//visualize = true;bool mutual_filter = false;//if (utility::ProgramOptionExists(argc, argv, "--mutual_filter")) {//    mutual_filter = true;//}mutual_filter = true;// Prepare inputstd::shared_ptr<geometry::PointCloud> source, target;std::shared_ptr<pipelines::registration::Feature> source_fpfh, target_fpfh;//set voxel_sizefloat voxel_size = std::atof(argv[3]);std::cout << voxel_size << std::endl;float distance_threshold = voxel_size * 1.5;std::tie(source, source_fpfh) = PreprocessPointCloud(argv[1], voxel_size);std::tie(target, target_fpfh) = PreprocessPointCloud(argv[2], voxel_size);pipelines::registration::RegistrationResult registration_result;// Prepare checkersstd::vector<std::reference_wrapper<const pipelines::registration::CorrespondenceChecker>>correspondence_checker;auto correspondence_checker_edge_length =pipelines::registration::CorrespondenceCheckerBasedOnEdgeLength(0.9);auto correspondence_checker_distance =pipelines::registration::CorrespondenceCheckerBasedOnDistance(distance_threshold);correspondence_checker.push_back(correspondence_checker_edge_length);correspondence_checker.push_back(correspondence_checker_distance);registration_result = pipelines::registration::RegistrationRANSACBasedOnFeatureMatching(*source, *target, *source_fpfh, *target_fpfh,mutual_filter, distance_threshold,pipelines::registration::TransformationEstimationPointToPoint(false),3, correspondence_checker,pipelines::registration::RANSACConvergenceCriteria(1000000, 0.999));cout << endl<< "fpfh matrix:" << endl<< registration_result.transformation_ << endl;cout << "inlier(correspondence_set size):"<< registration_result.correspondence_set_.size() << endl;std::vector<std::pair<int, int>> correspondences_ransac;for (int m = 0; m < registration_result.correspondence_set_.size();++m) {std::pair<int, int> pair_(0, 0);pair_.first = registration_result.correspondence_set_[m][0];pair_.second = registration_result.correspondence_set_[m][1];correspondences_ransac.push_back(pair_);}// std::shared_ptr<open3d::geometry::LineSet>auto ransac_lineset =geometry::LineSet::CreateFromPointCloudCorrespondences(*source, *target, correspondences_ransac);std::shared_ptr<geometry::PointCloud> source_transformed_ptr(new geometry::PointCloud);std::shared_ptr<geometry::PointCloud> source_ptr(new geometry::PointCloud);std::shared_ptr<geometry::PointCloud> target_ptr(new geometry::PointCloud);*source_transformed_ptr = *source;*target_ptr = *target;*source_ptr = *source;target->PaintUniformColor({1, 0, 0});                  //红source->PaintUniformColor({0, 1, 0});                  //绿source_transformed_ptr->PaintUniformColor({0, 0, 1});  //蓝source_transformed_ptr->Transform(registration_result.transformation_);if (visualize) {visualization::DrawGeometries({target, source, source_transformed_ptr, ransac_lineset},"Registration result", 960, 900, 960, 100);}return 0;
}

编译demo

cd open3d_fpfh_registration
mkdir build
cd build
cmake -DOpen3D_ROOT="D:/Program Files/open3d" ..
cmake --build . --config Debug

成功生成RegistrationRANSAC.exe
在这里插入图片描述

配准测试

数据就用斯坦福的兔兔,http://graphics.stanford.edu/pub/3Dscanrep/bunny.tar.gz
解压后把bun000.ply和bun045.ply放在open3d_fpfh_registration目录下
回到命令行窗口:

start ./Debug/RegistrationRANSAC.exe ../bun045.ply ../bun000.ply 0.005 --visualize

若是其他数据,改一下…/bun045.ply …/bun000.ply 0.005 --visualize这几个参数就好,分别是source点云路径,target点云路径,降采样网格大小(根据点云大小和数据量来自行调节),–visualize加可视化

配准结果,红色是bun000,绿色是bun045,蓝色是bun045配准至bun000的结果,连线是连接的匹配上的fpfh特征
在这里插入图片描述

参考

文中已列出

--------------------------------------------------------------------------------------------诺有缸的高飞鸟202110

这篇关于open3d使用fpfh做点云配准的文章就介绍到这儿,希望我们推荐的文章对编程师们有所帮助!



http://www.chinasem.cn/article/701183

相关文章

Python使用PIL库将PNG图片转换为ICO图标的示例代码

《Python使用PIL库将PNG图片转换为ICO图标的示例代码》在软件开发和网站设计中,ICO图标是一种常用的图像格式,特别适用于应用程序图标、网页收藏夹图标等场景,本文将介绍如何使用Python的... 目录引言准备工作代码解析实践操作结果展示结语引言在软件开发和网站设计中,ICO图标是一种常用的图像

使用Java发送邮件到QQ邮箱的完整指南

《使用Java发送邮件到QQ邮箱的完整指南》在现代软件开发中,邮件发送功能是一个常见的需求,无论是用户注册验证、密码重置,还是系统通知,邮件都是一种重要的通信方式,本文将详细介绍如何使用Java编写程... 目录引言1. 准备工作1.1 获取QQ邮箱的SMTP授权码1.2 添加JavaMail依赖2. 实现

MyBatis与其使用方法示例详解

《MyBatis与其使用方法示例详解》MyBatis是一个支持自定义SQL的持久层框架,通过XML文件实现SQL配置和数据映射,简化了JDBC代码的编写,本文给大家介绍MyBatis与其使用方法讲解,... 目录ORM缺优分析MyBATisMyBatis的工作流程MyBatis的基本使用环境准备MyBati

使用Python开发一个图像标注与OCR识别工具

《使用Python开发一个图像标注与OCR识别工具》:本文主要介绍一个使用Python开发的工具,允许用户在图像上进行矩形标注,使用OCR对标注区域进行文本识别,并将结果保存为Excel文件,感兴... 目录项目简介1. 图像加载与显示2. 矩形标注3. OCR识别4. 标注的保存与加载5. 裁剪与重置图像

使用Python实现表格字段智能去重

《使用Python实现表格字段智能去重》在数据分析和处理过程中,数据清洗是一个至关重要的步骤,其中字段去重是一个常见且关键的任务,下面我们看看如何使用Python进行表格字段智能去重吧... 目录一、引言二、数据重复问题的常见场景与影响三、python在数据清洗中的优势四、基于Python的表格字段智能去重

使用Apache POI在Java中实现Excel单元格的合并

《使用ApachePOI在Java中实现Excel单元格的合并》在日常工作中,Excel是一个不可或缺的工具,尤其是在处理大量数据时,本文将介绍如何使用ApachePOI库在Java中实现Excel... 目录工具类介绍工具类代码调用示例依赖配置总结在日常工作中,Excel 是一个不可或缺的工http://

Java之并行流(Parallel Stream)使用详解

《Java之并行流(ParallelStream)使用详解》Java并行流(ParallelStream)通过多线程并行处理集合数据,利用Fork/Join框架加速计算,适用于大规模数据集和计算密集... 目录Java并行流(Parallel Stream)1. 核心概念与原理2. 创建并行流的方式3. 适

如何使用Docker部署FTP和Nginx并通过HTTP访问FTP里的文件

《如何使用Docker部署FTP和Nginx并通过HTTP访问FTP里的文件》本文介绍了如何使用Docker部署FTP服务器和Nginx,并通过HTTP访问FTP中的文件,通过将FTP数据目录挂载到N... 目录docker部署FTP和Nginx并通过HTTP访问FTP里的文件1. 部署 FTP 服务器 (

MySQL 日期时间格式化函数 DATE_FORMAT() 的使用示例详解

《MySQL日期时间格式化函数DATE_FORMAT()的使用示例详解》`DATE_FORMAT()`是MySQL中用于格式化日期时间的函数,本文详细介绍了其语法、格式化字符串的含义以及常见日期... 目录一、DATE_FORMAT()语法二、格式化字符串详解三、常见日期时间格式组合四、业务场景五、总结一、

Python中配置文件的全面解析与使用

《Python中配置文件的全面解析与使用》在Python开发中,配置文件扮演着举足轻重的角色,它们允许开发者在不修改代码的情况下调整应用程序的行为,下面我们就来看看常见Python配置文件格式的使用吧... 目录一、INI配置文件二、YAML配置文件三、jsON配置文件四、TOML配置文件五、XML配置文件