CMake OpenCV4.2 SFM模块(Windows 10, VS2017)

    环境：Win10, OpenCV4.2，VS2017, CMake 3.9.1

    参考链接：https://blog.csdn.net/fanhao900322/article/details/78787712

                     https://blog.csdn.net/wzheng92/article/details/79504709 

    ① GFlags编译，源码链接：https://github.com/gflags/gflags/tree/v2.2.2

 

 

 

ALL_Build之后，Install仅用于项目生成，生成文件目录如下：

C:/Program Files/gflags/lib/gflags_static.lib
C:/Program Files/gflags/lib/gflags_nothreads_static.lib
C:/Program Files/gflags/include/gflags/gflags.h
C:/Program Files/gflags/include/gflags/gflags_declare.h
C:/Program Files/gflags/include/gflags/gflags_completions.h
C:/Program Files/gflags/include/gflags/gflags_gflags.h
C:/Program Files/gflags/lib/cmake/gflags/gflags-config.cmake
C:/Program Files/gflags/lib/cmake/gflags/gflags-config-version.cmake
C:/Program Files/gflags/lib/cmake/gflags/gflags-targets.cmake
C:/Program Files/gflags/lib/cmake/gflags/gflags-targets-release.cmake
C:/Program Files/gflags/lib/cmake/gflags/gflags-nonamespace-targets.cmake
C:/Program Files/gflags/lib/cmake/gflags/gflags-nonamespace-targets-release.cmake

    ② glog编译：源码链接：https://github.com/google/glog/tree/v0.4.0

    需要依赖上一步生成的GFlags Install目录，然后同样VS2017生成，Install生成文件目录：

C:/Program Files/glog/lib/glog.lib
C:/Program Files/glog/include/glog/logging.h
C:/Program Files/glog/include/glog/raw_logging.h
C:/Program Files/glog/include/glog/stl_logging.h
C:/Program Files/glog/include/glog/vlog_is_on.h
C:/Program Files/glog/include/glog/log_severity.h
C:/Program Files/glog/lib/cmake/glog/glog-config.cmake
C:/Program Files/glog/lib/cmake/glog/glog-config-version.cmake
C:/Program Files/glog/lib/cmake/glog/glog-targets.cmake
C:/Program Files/glog/lib/cmake/glog/glog-targets-release.cmake

    ③ Eigen3.3.7编译，源码链接：https://gitlab.com/libeigen/eigen/-/releases

ALL_Build之后，Install仅用于项目生成，生成文件目录C:/Program Files/Eigen3/include/eigen3...

    ④ Suitesparse编译，源码链接：https://github.com/jlblancoc/suitesparse-metis-for-windows/tree/v1.5.0

    ⑤ Ceres-solver编译，链接https://ceres-solver.googlesource.com/ceres-solver/+/refs/tags/1.13.0

        点击【tgz】下载，下载后需要解压两层才能获得Ceres源代码

    ⑥ OpenCV编译，需要配置前面几个路径，可以参考以下三个链接，相信你可以搞定了

         https://blog.csdn.net/wzheng92/article/details/79504709

         https://www.jianshu.com/p/89a5705f0941    

         https://blog.csdn.net/fanhao900322/article/details/78787712

     

    使用的时候注意添加opencv_sfm420.lib

    还可能会遇到一个问题，就是提示reconstruct函数无法识别的标识符，解决方法在#include 前

添加#define CERES_FOUND 1

    接下来运行一个实例，代码见官方文档https://docs.opencv.org/master/d4/d18/tutorial_sfm_scene_reconstruction.html

注意运行时将前面生成的dll都复制过来或者加到环境变量(推荐)

运行生成的点云感觉很稀疏，和官方结果不一样，需要再研究研究问题在哪