CMake编译好OpenMVG然后运行tutorial_demo.py，报错如下:

CleanGraph_KeepLargestBiEdge_Nodes():: => connected Component: 0
Invalid input image graph for global SfM
5. Colorize Structure

The input SfM_Data file "D:\OpenCV_SFM_Dependency\openMVG\build\software\SfM\tutorial_out\reconstruction_global/sfm_data.bin" cannot be read.
4. Structure from Known Poses (robust triangulation)
Compute Structure from the provided poses

The input SfM_Data file "D:\OpenCV_SFM_Dependency\openMVG\build\software\SfM\tutorial_out\reconstruction_global/sfm_data.bin" cannot be read.

问题是我们没有设置相机焦距参数，需要修改一下tutorial_demo.py的这部分代码：

修改前：

pIntrisics = subprocess.Popen( [os.path.join(OPENMVG_SFM_BIN, "openMVG_main_SfMInit_ImageListing"),
                                 "-i", input_dir, "-o", matches_dir, "-d", camera_file_params, "-c", "3"] )

修改后：

pIntrisics = subprocess.Popen( [os.path.join(OPENMVG_SFM_BIN, "openMVG_main_SfMInit_ImageListing"), "-i", input_dir,\
                                "-o", matches_dir, "-d", camera_file_params, "-f", "1228.8"]);

注意这里焦距f的设置github说明 f = max(Width, Height) * 1.2，也就是图像宽高中的最大值乘以1.2，我的图像宽高是1024*768，

所以1024 * 1.2 = 1228.8

更改之后，设置直接的重建图片路径，就可以生成sfm_data.bin，接下来就参考这篇文章继续往下做吧https://blog.csdn.net/X_kh_2001/article/details/83690094

https://github.com/cdcseacave/openMVS/wiki/Usage