基于openMVG和openMVS的实用编程Python+Pyqt5

基于openMVG和openMVS的使用编程Python+Pyqt5

前言：对于许多要建立三维模型的小伙伴而言，openMVG+openMVS是目前不多的开源，可用的资源，时隔半年，我又来搞了。

本次使用教程使用Python3.8+Pyqt5
（ps：python真的太好用了，QT真的太好用啦，完全停不下来，C++/MFC：所以爱会消失对吗？ 对的）

---

1.由配置好的openMVG+openMVS编写程序

(1)我就是用配置好的exe写的，因为简单方便。
/1/ .如果没有配出来的小伙伴也可以不用配置，直接下载已经配置好的文件，如果有需要，邮箱联系[email protected]或私信
/2/ .目前GitHub上也有大佬用源代码写的，参考网址链接: link
注：大部分地区进入GitHub需要VPN哦
本次程序所有所需头文件如下，为了增强阅读体验，已删除QT界面以及线程部分代码

import sys
import win32gui
import win32con
import subprocess
import time
import os
import shutil

# from Show3D import Ui_MainWindow
from PyQt5.QtWidgets import QWidget, QApplication, QMessageBox, QLabel, QProgressBar
from PyQt5.QtGui import QWindow
from PyQt5 import QtCore, QtWidgets
from PyQt5.QtCore import QThread, QMutex, pyqtSignal

(2)Python+openMVG
其实这部分在配置openMVG时，实例文件可以查到，在windows下使用方法如下：
**exe -i inputfile -o outputfile -m method **
解释：运行exe， 传入参数，-i表示输入，-o表示输出，等等，openMVG完整代码如下

		# 路径设定
        openMVG_dir = 'D:/show3D/openMVG/'
        openMVS_dir = 'D:/show3D/openMVS/bin/'
        images_dir = 'D:/show3D/images/'
        # 相机参数文件，在openMVG源文件下
        CAMERA_SENSOR_WIDTH_DIRECTORY = 'D:/show3D/openMVG/sensor_width_camera_database.txt'
        openMVG_output_dir = 'D:/show3D/openMVG_output/'
        openMVS_output_dir = 'D:/show3D/openMVS_output/'
        # openMVG步骤
        # 清空上次运行结果
        shutil.rmtree(openMVG_output_dir)
        os.mkdir(openMVG_output_dir)
        psubprocess = subprocess.Popen(
            [os.path.join(openMVG_dir, "openMVG_main_SfMInit_ImageListing"), "-i", images_dir,
             "-d", CAMERA_SENSOR_WIDTH_DIRECTORY, "-o", openMVG_output_dir])
        psubprocess.wait()
        psubprocess = subprocess.Popen(
            [os.path.join(openMVG_dir, "openMVG_main_ComputeFeatures"), "-i", openMVG_output_dir + "sfm_data.json",
             "-o", openMVG_output_dir, "-m", "SIFT", "-p", "HIGH"])

        psubprocess.wait()

        psubprocess = subprocess.Popen([os.path.join(openMVG_dir, "openMVG_main_ComputeMatches"),
                                        "-i", openMVG_output_dir + "sfm_data.json", "-o", openMVG_output_dir])
        psubprocess.wait()

        psubprocess = subprocess.Popen([os.path.join(openMVG_dir, "openMVG_main_IncrementalSfM"),
                                        "-i", openMVG_output_dir + "sfm_data.json", "-m", openMVG_output_dir,
                                        "-o", openMVG_output_dir])
        psubprocess.wait()
        psubprocess = subprocess.Popen([os.path.join(openMVG_dir, "openMVG_main_ComputeSfM_DataColor"),
                                        "-i", openMVG_output_dir + "sfm_data.bin",
                                        "-o", openMVG_output_dir + "colored.ply"])
        psubprocess.wait()

        psubprocess = subprocess.Popen([os.path.join(openMVG_dir, "openMVG_main_ComputeStructureFromKnownPoses"),
                                        "-i", openMVG_output_dir + "sfm_data.bin", "-m", openMVG_output_dir,
                                        "-o", openMVG_output_dir + "robust.bin",
                                        "-f", openMVG_output_dir + "matches.f.bin"])
        psubprocess.wait()

        psubprocess = subprocess.Popen([os.path.join(openMVG_dir, "openMVG_main_openMVG2openMVS"),
                                        "-i", openMVG_output_dir + "sfm_data.bin",
                                        "-o", openMVG_output_dir + "scene.mvs"])
        psubprocess.wait()

注：建议每次调用时，清空上次运行的结果，亲测，上次文件仍然在，会发生各种意外的错误
（2）openMVS步骤
其实openMVS也可以用openMVG的方法调用，只不过在python中不显示运行结果，具体代码如下：

        # openMVS步骤
        shutil.rmtree(openMVS_output_dir)
        os.mkdir(openMVS_output_dir)
        shutil.copy2(openMVG_output_dir + "scene.mvs", openMVS_output_dir + "scene.mvs")
        psubprocess = subprocess.Popen([os.path.join(openMVS_dir, "DensifyPointCloud.exe"),
                                        "-i", os.path.join(openMVS_output_dir, 'scene.mvs')])
        psubprocess.wait()

        psubprocess = subprocess.Popen([os.path.join(openMVS_dir, "ReconstructMesh.exe"),
                                        "-i", os.path.join(openMVS_output_dir, "scene_dense.mvs")])
        psubprocess.wait()

        psubprocess = subprocess.Popen([os.path.join(openMVS_dir, "RefineMesh.exe"),
                                        "-i", os.path.join(openMVS_output_dir, "scene_dense_mesh.mvs")])
        psubprocess.wait()

        psubprocess = subprocess.Popen([os.path.join(openMVS_dir, "TextureMesh.exe"),
                                        "-i", os.path.join(openMVS_output_dir, "scene_dense_mesh_refine.mvs")])
        psubprocess.wait()

        self.qmut.unlock()
        self.DeleteFile()
   # 删除根目录下生成的文件，主要都是.log和.dmap
   def DeleteFile(self):
        file_path = 'D:/pythonProject/sfm/'
        for roots, dirs, files in os.walk(file_path):
            for file in files:
                if file.endswith('.log') or file.endswith('.dmap'):
                    os.remove(os.path.join(roots, file))

注：Windows下编译的openMVS不带Viewer，有需要的小伙伴私信，邮箱发配置好的整套openMVS

2.总结

以上就是所有运行openMVG和openMVS的代码。查看最后生成的mvs文件使用编译的Viewer.exe。cmd命令为：Viewer.exe scene_dense_mesh_refine_texture.mvs。python命令中间加 -i

下一章，将Viewer显示的图像内嵌到QT窗口中，并保留旋转等功能。祝大家心想事成。