8

1. 前言

鉴于VS本身体积的庞大和项目依赖管理方式的不便,所以本文采用Clion通过CMake进行项目结构管理。

2. 准备工作

  • Windows 7 + (本文环境为 win7 x64)
  • Clion
  • MinGW(本文环境 tdm-gcc-5.1.0-3)
  • CMake (本文环境 cmake-3.7.2)
  • OpenCV(本文环境 opencv-3.4.1)

安装以上软件,基本都是一路 next ,最后做一下系统环境变量的配置(右击我点电脑->属性->高级系统设置->环境变量->系统变量->编辑Path,添加对应软件bin所在目录,中间分号分隔)。最后打开Clion,依次进入Setting->Build,Execution,Deployment->Toolchains,设置编译用工具链:

工具链设置

ps:编译好的OpenCV MSVC版及 MinGW版

  • MSVC版

    下载的 OpenCV 文件夹会有:

    • build (已编译好的库)
    • sources (源码)
使用 MSVC 的话,直接在

    build/x64/vc14

里面就有了,配置好路径即可使用
  • MinGW版

    OpenCV 没有为我们编译好 MinGW 版,所以只能用户自行编译,下面就是介绍 MinGW 版的编译流程以及中间可能遇到的错误的排除

    这里有需要的话可以直接下载使用我编译好了的 OpenCV-3.4.1 x64

3. 编译

  1. 解压OpenCV,然后在解压后source同级目录下创建一个新的mingw-build文件夹用于存放编译后的内容。

    (ps:这里如果你之前安装了Anaconda或者Python,请把这两个软件暂时从环境变量中删除,只是删除系统Path中的存在即可,不是卸载软件。原因是后面编译会有冲突)

  2. 然后这里选择用cmake图形化界面操作,比较直观一些,采用命令行注意下命令参数同理。source和binary目录按如下图选择:

    cmake

    其中:source目录对应你的opencv/sources目录,binary目录就是之前新建的那个mingw-build存放目录

    此处需要复制opencv->build->bin下的两个文件:opencv_ffmpegxxx.dll、opencv_ffmpegxxx_64.dll到opencv/sources/3rdparty/ffmpeg/目录下

  3. 选好目录,点击configure选择选择MinGW Makefiles,如上图所示。

注意:遇见红色之后再次点击Configure(等于是要点两次),等到所有列表变白,没有红色一片的时候才表示成功

  1. configure成功后点击generate
  2. 进入刚才创建的mingw-build目录,直接右键Git bash here(如果没装git windows,直接在cmd下进行),然后使用make命令进行编译:

    mingw32-make -j2 # 以2线程进行编译,这个数字根据自己硬件合理选择

    make_install

  3. 漫长的make到达100%后进行install:

    mingw32-make install

  4. 添加环境变量

    • 为系统变量 Path 添加 C:opencvmingw-buildinstallx64mingwbin
    • 添加 OpenCV_DIR (注意大小写,部分情况这个可以不加,不过我这里不加Cmake会报找不到OpenCV的错误),变量值为之前编译的build路径:C:opencvmingw-build (实质是告知Cmake 包含OpenCV库文件和include文件配置文件的OpenCVConfig.cmake所在)

opencv_dir

4. 调用

至此,OpenCV编译和相关设置已经完成,只需要正常调用即可使用。下面给一个简单摄像头调用示例:

CMakeLists.txt

cmake_minimum_required(VERSION 3.0)
project(detect)

set(CMAKE_CXX_STANDARD 11)

add_executable(detect main.cpp)

FIND_PACKAGE(OpenCV REQUIRED)
IF (OpenCV_FOUND)
    INCLUDE_DIRECTORIES(${OpenCV_INCLUDE_DIRS})
    TARGET_LINK_LIBRARIES(detect ${OpenCV_LIBS})
ELSE (OpenCV_FOUND)
    MESSAGE(FATAL_ERROR "OpenCV library not found")
ENDIF (OpenCV_FOUND)

main.cpp

#include <opencv2/highgui.hpp>
#include <opencv2/imgproc.hpp>
#include <iostream>

int main() {
cv::VideoCapture capture(0);

if (!capture.isOpened()) {
    std::cout << "open camera error!" << std::endl;
    return -1;
}

cv::Mat frame;
while (1) {
    capture >> frame;
    if (frame.empty()) {
        std::cout << "capture empty frame" << std::endl;
        continue;
    }

    cv::Mat shrink_frame;
    cv::resize(frame, shrink_frame,
               cv::Size(frame.cols / 2, frame.rows / 2),
               0, 0, 3);

    cv::imshow("detect", shrink_frame);

    int key = cv::waitKey(1);
    if (key == 'q') {
        break;
    }
}
return 0;
}

detect

5. 编译常见错误排除

  • 多线程编译错误信息不明确

    表现:

    如果使用了多线程编译,导致错误,但是错误信息不明确,如:

     Makefile:161: recipe for target 'all' failed
     mingw32-make: *** [all] Error 2
     

    多线程错误

    解决:

    使用单线程make,以查看详细的错误提示,再根据具体情况解决

     mingw32-make
     
  • RC 错误

    表现:

     ... windres.exe: unknown option -- W ...
    

    或者:

     
     FORMAT is one of rc, res, or coff, and is deduced from the file name
     extension if not specified.  A single file name is an input file.
     No input-file is stdin, default rc.  No output-file is stdout, default rc.
     

    RC错误

    解决:

    在 cmake-gui 编译配置中:不勾选 ENABLE PRECOMPILED HEADERS,然后重新Configure-Generate-mingw32-make

  • sprintf instead use StringCbPrintfA or StringCchPrintfA 错误

    表现:

     ...opencv/sources/modules/videoio/src/cap_dshow.cpp...
     ... 'sprintf_instead_use_StringCbPrintfA_or_StringCchPrintfA' was not declared in this scope ...
     

    sprintf错误

    解决:

    修改E:OpenCV_3.3.1opencvsourcesmodulesvideoiosrccap_dshow.cpp文件,在#include "DShow.h"这行的上面加一行#define NO_DSHOW_STRSAFE,如:

     #define NO_DSHOW_STRSAFE
     #include "DShow.h"
     

    然后重新Configure-Generate-mingw32-make

  • identifier ‘nullptr’ is a keyword in C++11 错误

    表现:

     D:\opencv-3.4.1\opencv-3.4.1\3rdparty\protobuf\src\google\protobuf\stubs\io_win32.cc:94:3: warning: identifier 'nullptr' is a keyword in C++11 [-Wc++0x-compat]
     return s == nullptr || *s == 0;
    
     D:\opencv-3.4.1\opencv-3.4.1\3rdparty\protobuf\src\google\protobuf\stubs\io_win32.cc: In function 'bool google::protobuf::internal::win32::{anonymous}::null_or_empty(const char_type*)':
     D:\opencv-3.4.1\opencv-3.4.1\3rdparty\protobuf\src\google\protobuf\stubs\io_win32.cc:94:15: error: 'nullptr' was not declared in this scope
        return s == nullptr || *s == 0;
            ^
     3rdparty\protobuf\CMakeFiles\libprotobuf.dir\build.make:412: recipe for target '3rdparty/protobuf/CMakeFiles/libprotobuf.dir/src/google/protobuf/stubs/io_win32.cc.obj' failed
     mingw32-make[2]: *** [3rdparty/protobuf/CMakeFiles/libprotobuf.dir/src/google/protobuf/stubs/io_win32.cc.obj] Error 1
     CMakeFiles\Makefile2:710: recipe for target '3rdparty/protobuf/CMakeFiles/libprotobuf.dir/all' failed
     mingw32-make[1]: *** [3rdparty/protobuf/CMakeFiles/libprotobuf.dir/all] Error 2
     Makefile:161: recipe for target 'all' failed
     mingw32-make: *** [all] Error 2        

    KEY错误

    解决:

    在 cmake-gui 编译配置中:

    勾选 ENABLE_CXX11

    然后重新Configure-Generate-mingw32-make

  • dnn build error (Assembler messages)

    表现:

     C:\Users\ADMINI~1\AppData\Local\Temp\ccHIQxbw.s: Assembler messages:
     C:\Users\ADMINI~1\AppData\Local\Temp\ccHIQxbw.s:21800: Error: invalid register for .seh_savexmm
     ...
     modules\dnn\CMakeFiles\opencv_dnn.dir\build.make:1741: recipe for target 'modules/dnn/CMakeFiles/opencv_dnn.dir/layers/layers_common.avx512_skx.cpp.obj' failed
     mingw32-make[2]: *** [modules/dnn/CMakeFiles/opencv_dnn.dir/layers/layers_common.avx512_skx.cpp.obj] Error 1
     CMakeFiles\Makefile2:4124: recipe for target 'modules/dnn/CMakeFiles/opencv_dnn.dir/all' failed
     mingw32-make[1]: *** [modules/dnn/CMakeFiles/opencv_dnn.dir/all] Error 2
     Makefile:159: recipe for target 'all' failed
     mingw32-make: *** [all] Error 2
     

    dnn error

    解决:

    在 cmake-gui 编译配置中:

    DCPU_DISPATCH 选空白

    然后重新Configure-Generate-mingw32-make

    (ps:这个错误在最新的OpenCV master分支已修复)

ref:

dnn build error via Mingw64

OpenCV使用CMake和MinGW的编译安装


ISherry
902 声望35 粉丝

Coding...