windows：Opencv使用Cmake&VS编译，C++、python、Cuda、Qt环境详细教学及踩坑一：编译篇

见贤思齐

求赞求赞求赞，若有帮助，鼓励一下测试篇：windows：Opencv使用Cmake&VS编译，C++、python、Cuda、Qt环境详细教学及踩坑二：测试篇-CSDN博客一、准备：本文使用的软件版本均是撰写时的最新或次新版，对过旧版本不一定通用0、环境配置：WindowsWindows11专业版 23H2opencv4.10Qt  6.7.2Cmake3.3.0python3.12.4cuda12.5cudnn9.2.1vsstudio20221、Opencv，Opencv_contrib 源码下载   本文版本：4.10GitHub-opencv/opencv:OpenSourceComputerVisionLibraryGitHub-opencv/opencv_contrib:RepositoryforOpenCV'sextramodules下载完成后解压opencv和opencv_contrib，将opencv_contrib文件夹移入opencv文件夹下，并在opencv文件夹下新建build文件夹。 软件安装教程引用其他大神的教程，很详细2、Qt安装          本文版本：Qt6.7.2Qt下载安装及配置教程(非常详细)，从零基础入门到精通，看完这一篇就够了-CSDN博客3、Cmake安装：    本文版本：Cmake3.3.0Windows下的CMake详细安装教程（2024/1/24）_windows安装cmake-CSDN博客4、python安装：        本文版本：python3.12.4 DownloadPython|Python.org1、选择自定义安装2、选择添加python路径到环境变量3、选择为所有用户安装其他选项如图所示，如果cmake需要debug版本的一定要勾选debug有关的选项，否则生成过程中会缺少python3xx_d.lib 文件。   5、Cuda、cudNN安装：本文版本：Cuda 12.5   cudNN9.2 先按该教程安装cuda、cudnn：Cuda和cuDNN安装教程(超级详细)-CSDN博客 cudnn再按官方文档安装一遍：英伟达官方cudnn安装教程 注意：cudnn一定要按教程，将cudnn文件复制到cuda文件夹下。否则：如果只是按照官方文档中的教程，编译成功后，pythonimportcv2时无法调用cudnn中的dll文件。报错：ImportErrorLLloadfailed:Thespecifiedmodulecouldnotbefound.！！注意：Windows下使用cmake编译cuda只能使用VS，否则会无法找到cuda位置6、VSstudio安装：    ThankYouforDownloadingVisualStudioCommunityEdition7、上述软件安装完成后重启计算机！！！二、设置Cmake选项    参考：AMDCustomerCommunity从“开始”菜单打开CMake（cmake-gui）将“Whereisthesourcecode”设置为提取的opencv文件夹。将“Wheretobuildthebinaries”设置为提取的opencv文件夹，并在末尾附加/build。这将是编译文件的构建位置，您可以将其设置为所需的任何文件夹点击“configure”弹出窗口选择vs2019作为编译器generate平台选择x64注：此处 opencv_cp312_410、opencv_410_cp312_amd64_bulid、opencv_410_cp312_amd64为笔者按习惯重命名。对应关系为：                opencv_cp312_410——>opencv_x.xx  opencv主文件夹                opencv_410_cp312_amd64_bulid——>build        opencv_410_cp312_amd64  ——>install此处的目录结构为：----opencv_x.xx    ----3rdparty       ----.....    ----build      ----install    ----.....       ----opencv_contrib​1、通用编译选项：参考链接：AMDCustomerCommunity        使用“搜索”框更改以下属性：BUILD_PROTOBUF：–取消选中WITH_PROTOBUF：–取消选中BUILD_TESTS：–取消选中WITH_OPENEXR：–取消选中BUILD_OPENEXR：–取消选中OPENCV_ENABLE_ALLOCATOR_STATS：–取消选中CMAKE_BUILD_TYPE：–RELEASECMAKE_INSTALL_PREFIX：–这默认为build/install目录，但您可以将其更改为您希望将OpenCV库安装到的任何目录。BUILD_opencv_world：建议勾选，build_opencv_world可以将所有opencv的库都打包编译在一起。OPENCV_ENABLE_NONFREE：勾选，否则无法使用sift。WITH_OPENGL ：勾选OPENCV_EXTRA_MODULES_PATH –设置为提取的 opencv_contrib/modules文件夹的路径。如图所示：        10、配置完成点击“configure”报错信息1：CMakeWarning(dev)atCMakeLists.txt:127(enable_language): project()shouldbecalledpriortothisenable_language()call.Thiswarningisforprojectdevelopers. Use-Wno-devtosuppressit.SelectingWindowsSDKversion10.0.22621.0totargetWindows10.0.22631.CMakeWarning(dev)atC:/ProgramFiles/CMake/share/cmake-3.30/Modules/Platform/Windows-MSVC.cmake:539(enable_language): project()shouldbecalledpriortothisenable_language()call.CallStack(mostrecentcallfirst): C:/ProgramFiles/CMake/share/cmake-3.30/Modules/Platform/Windows-MSVC.cmake:509(__windows_compiler_msvc_enable_rc) C:/ProgramFiles/CMake/share/cmake-3.30/Modules/Platform/Windows-MSVC-CXX.cmake:6(__windows_compiler_msvc) C:/ProgramFiles/CMake/share/cmake-3.30/Modules/CMakeCXXInformation.cmake:48(include) CMakeLists.txt:127(enable_language)Thiswarningisforprojectdevelopers. Use-Wno-devtosuppressit.报错信息2： CMakeWarningatcmake/OpenCVGenSetupVars.cmake:54(message): CONFIGURATIONISNOTSUPPORTED:validatesetupvarsscriptininstall directoryCallStack(mostrecentcallfirst): CMakeLists.txt:1180(include)​解决方案： 这两类警告，不影响正常编译过程。对此未作处理。2、cuda编译选项参考链接：windows11编译OpenCV4.5.0withCUDA（附注意事项）-CSDN博客第一次搜cuda时将带cuda的都勾选，点击“configure”搜索：WITH_NVCUVENC——取消选中      WITH_NVCUVID——取消选中注意：第一次搜索时，如图所示的部分选项不存在是正常现象，因为每次"configure"时，cmake都会根据已经勾选的选项重新加载，所以及时的重新"configure"是一个好习惯。将已经存在的选项按图勾选后，再次点击"configure"后重新搜索，即可出现全部选项。 配置CUDA_ARCH_BIN中将显卡的算力内容改成自己显卡的算力,显卡算力查看链接查看自己显卡的算力，我的显卡是4060laptop，官网算力为8.9。CUDA_ARCH_BIN中将低于8.9数值的删除。​cuda编译选项按图进行勾选。​报错信息1：WITH_NVCUVENC、WITH_NVCUVIDCMakeWarningatopencv_contrib_4_10/modules/cudacodec/CMakeLists.txt:26(message): cudacodec::VideoReaderrequiresNvidiaVideoCodecSDK. leaseresolve dependencyordisableWITH_NVCUVID=OFFCallStack(mostrecentcallfirst): modules/world/CMakeLists.txt:13(include) modules/world/CMakeLists.txt:50(include_one_module)CMakeWarningatopencv_contrib_4_10/modules/cudacodec/CMakeLists.txt:30(message): cudacodec::VideoWriterrequiresNvidiaVideoCodecSDK. leaseresolve dependencyordisableWITH_NVCUVENC=OFFCallStack(mostrecentcallfirst): modules/world/CMakeLists.txt:13(include) modules/world/CMakeLists.txt:50(include_one_module)解决方法：               搜索：WITH_NVCUVENC——取消选中          WITH_NVCUVID——取消选中报错信息2：OPENCV_DNN_CUDACMakeErroratmodules/dnn/CMakeLists.txt:53(message): DNN:CUDAbackendrequirescuDNN. leaseresolvedependencyordisable OPENCV_DNN_CUDA=OFFCallStack(mostrecentcallfirst): modules/world/CMakeLists.txt:13(include) modules/world/CMakeLists.txt:50(include_one_module)​解决方法：搜索CUDNN_LIBRARY，找到并输入cudnn.lib的安装位置​CUDA参考配置： 打开Advance，配置供参考（不一定要一致，只是提供参考）​3、python编译选项参考：【opencv】【GPU】windows10下opencv4.8.0-cudaPython版本源码编译教程_cv2的版本4.8.0-CSDN博客注意：有的同学即使按图勾选后依然无法出现某些选项。例如：    若无法出现BUILD_opencv_python3，（此选项为python编译的必需选项）此时可以进行手动添加。点击AddEntry ，在Name中输入BUILD_opencv_python3，点击OK，即可出现该选项。python的选项设置如图所示，其中numpy需提前进行安装，安装完成python后，打开cmd，输入pipinstallnumpy​​​python参考配置：4、Qt编译选项 参考：基于QT5.14.2的OPENCV4.8.0的cuda11.4开发环境搭建和源码编译_opencv4.8编译-CSDN博客 1、搜索Qt，设置Qt6_DIR目录为：Qt安装路径/6.7.2/msvc2019_64/lib/cmake/Qt6，如图所示。点击“configure”。  2、查找Qt安装路径\6.7.2\msvc2019_64\lib\cmake\Qtest文件夹，将其压缩备份后删除。 参考：https://github.com/opencv/opencv/issues/23826#issuecomment-2227167370否则会报错。点击“configure”。此时搜索 Qt6Test_DIR 为 Qt6Test_DIR-NOTFOUND。 报错信息：QtestCMakeErroratC:/ProgramFiles/CMake/share/cmake-3.30/Modules/FindCUDA.cmake:1793(file):Errorevaluatinggeneratorexpression$mayonlybeusedwithbinarytargets.Itmaynotbeusedwithadd_custom_commandoradd_custom_target.Specifythetargettoreadapropertyfromusingthe$signatureinstead.CallStack(mostrecentcallfirst):C:/ProgramFiles/CMake/share/cmake-3.30/Modules/FindCUDA.cmake:2111(CUDA_WRAP_SRCS)C:/ProgramFiles/CMake/share/cmake-3.30/Modules/FindCUDA.cmake:2124(cuda_compile_base)cmake/OpenCVDetectCUDA.cmake:164(CUDA_COMPILE)cmake/OpenCVUtils.cmake:1578(ocv_cuda_compile)cmake/OpenCVModule.cmake:989(ocv_add_library)cmake/OpenCVModule.cmake:905(_ocv_create_module)modules/world/CMakeLists.txt:78(ocv_create_module)解决方案： 查找Qt安装路径\6.7.2\msvc2019_64\lib\cmake\Qtest文件夹，将其压缩备份后删除。 参考：https://github.com/opencv/opencv/issues/23826#issuecomment-2227167370Qt参考配置：注意：每次点击"configure"都要注意报错信息，除了一开始的两个警告外，其他警告一定要及时解决，否则编译会失败三、检查配置报告 ！！！检查具有背景色的行，看是否与自己预期一致。GeneralconfigurationforOpenCV4.10.0=====================================Versioncontrol:unknownExtramodulesocation(extra):/ProgramFiles/opencv/opencv_cp312_410/opencv_contrib_4_10/modulesVersioncontrol(extra):unknownPlatform:Timestamp:2024-07-19T05:23:16ZHost:Windows10.0.22631AMD64CMake:3.30.0CMakegenerator:VisualStudio172022CMakebuildtool:C:/ProgramFiles(x86)/MicrosoftVisualStudio/2022/BuildTools/MSBuild/Current/Bin/amd64/MSBuild.exeMSVC:1940ConfigurationebugReleaseCPU/HWfeatures:Baseline:SSESSE2SSE3requested:SSE3Dispatchedcodegeneration:SSE4_1SSE4_2FP16AVXAVX2AVX512_SKXrequested:SSE4_1SSE4_2AVXFP16AVX2AVX512_SKXSSE4_1(16files):+SSSE3SSE4_1SSE4_2(1files):+SSSE3SSE4_1POPCNTSSE4_2FP16(0files):+SSSE3SSE4_1POPCNTSSE4_2FP16AVXAVX(8files):+SSSE3SSE4_1POPCNTSSE4_2AVXAVX2(36files):+SSSE3SSE4_1POPCNTSSE4_2FP16FMA3AVXAVX2AVX512_SKX(5files):+SSSE3SSE4_1POPCNTSSE4_2FP16FMA3AVXAVX2AVX_512FAVX512_COMMONAVX512_SKXC/C++:Builtasdynamiclibs?:YESC++standard:11C++Compiler:C:/ProgramFiles(x86)/MicrosoftVisualStudio/2022/BuildTools/VC/Tools/MSVC/14.40.33807/bin/Hostx64/x64/cl.exe(ver19.40.33812.0)C++flags(Release):/DWIN32/D_WINDOWS/W4/GR/D_CRT_SECURE_NO_DEPRECATE/D_CRT_NONSTDC_NO_DEPRECATE/D_SCL_SECURE_NO_WARNINGS/Gy/bigobj/Oi/fp:precise/EHa/wd4127/wd4251/wd4324/wd4275/wd4512/wd4589/wd4819/MP/O2/Ob2/DNDEBUGC++flags(Debug):/DWIN32/D_WINDOWS/W4/GR/D_CRT_SECURE_NO_DEPRECATE/D_CRT_NONSTDC_NO_DEPRECATE/D_SCL_SECURE_NO_WARNINGS/Gy/bigobj/Oi/fp:precise/EHa/wd4127/wd4251/wd4324/wd4275/wd4512/wd4589/wd4819/MP/Zi/Ob0/Od/RTC1CCompiler:C:/ProgramFiles(x86)/MicrosoftVisualStudio/2022/BuildTools/VC/Tools/MSVC/14.40.33807/bin/Hostx64/x64/cl.exeCflags(Release):/DWIN32/D_WINDOWS/W3/D_CRT_SECURE_NO_DEPRECATE/D_CRT_NONSTDC_NO_DEPRECATE/D_SCL_SECURE_NO_WARNINGS/Gy/bigobj/Oi/fp:precise/MP/O2/Ob2/DNDEBUGCflags(Debug):/DWIN32/D_WINDOWS/W3/D_CRT_SECURE_NO_DEPRECATE/D_CRT_NONSTDC_NO_DEPRECATE/D_SCL_SECURE_NO_WARNINGS/Gy/bigobj/Oi/fp:precise/MP/Zi/Ob0/Od/RTC1Linkerflags(Release):/machine:x64/INCREMENTAL:NOLinkerflags(Debug):/machine:x64/debug/INCREMENTALccache:NOPrecompiledheaders:NOExtradependencies:cudart_static.libnppc.libnppial.libnppicc.libnppidei.libnppif.libnppig.libnppim.libnppist.libnppisu.libnppitc.libnpps.libcublas.libcudnn.libcufft.lib-LIBPATH:C:/ProgramFiles/NVIDIAGPUComputingToolkit/CUDA/v12.5/lib/x64-LIBPATH:C:/ProgramFiles/NVIDIAGPUComputingToolkit/CUDNN_9.2/lib/x643rdpartydependencies:OpenCVmodules:Tobebuilt:arucobgsegmbioinspiredcalib3dccalibcorecudaarithmcudabgsegmcudacodeccudafeatures2dcudafilterscudaimgproccudalegacycudaobjdetectcudaoptflowcudastereocudawarpingcudevdatasetsdnndnn_objdetectdnn_superresdpmfacefeatures2dflannfuzzygapihfshighguiimg_hashimgcodecsimgprocintensity_transformline_descriptormccmlobjdetectoptflowphase_unwrappingphotoplotpython3qualityrapidregrgbdsaliencyshapesignalstereostitchingstructured_lightsuperressurface_matchingtexttrackingtsvideovideoiovideostabwechat_qrcodeworldxfeatures2dximgprocxobjdetectxphotoDisabled:-Disabledbydependency:-Unavailable:alphamatcannopscvvfreetypehdfjavajuliamatlabovispython2python2sfmvizApplications:perf_testsappsDocumentation:NONon-freealgorithms:YESWindowsRTsupport:NOGUIT:NOWin32UI:YESOpenGLsupport:YES(opengl32glu32)VTKsupport:NOMediaI/O:ZLib:build(ver1.3.1)JPEG:build-libjpeg-turbo(ver3.0.3-70)SIMDSupportRequest:YESSIMDSupport:NOWEBP:build(verencoder:0x020f)PNG:build(ver1.6.43)SIMDSupportRequest:YESSIMDSupport:YES(IntelSSE)TIFF:build(ver42-4.6.0)JPEG2000:build(ver2.5.0)OpenEXR:build(ver2.3.0)HDR:YESSUNRASTER:YESPXM:YESPFM:YESVideoI/OC1394:NOFFMPEG:YES(prebuiltbinaries)avcodec:YES(58.134.100)avformat:YES(58.76.100)avutil:YES(56.70.100)swscale:YES(5.9.100)avresample:YES(4.0.0)GStreamer:NODirectShow:YESMediaFoundation:YESDXVA:YESParallelframework:ConcurrencyTrace:YES(withIntelITT)Otherthird-partylibraries:IntelIPP:2021.11.0[2021.11.0]at:/ProgramFiles/opencv/opencv_cp312_410/opencv_410_cp312_amd64_bulid/3rdparty/ippicv/ippicv_win/icvIntelIPPIW:sources(2021.11.0)at:/ProgramFiles/opencv/opencv_cp312_410/opencv_410_cp312_amd64_bulid/3rdparty/ippicv/ippicv_win/iwLapack:NOEigen:NOCustomHAL:NOFlatbuffers:builtin/3rdparty(23.5.9)NVIDIACUDA:YES(ver12.5,CUFFTCUBLAS)NVIDIAGPUarch:8990NVIDIAPTXarchs:90cuDNN:YES(ver9.2.1)OpenCL:YES(NVD3D11)Includepath:/ProgramFiles/opencv/opencv_cp312_410/3rdparty/include/opencl/1.2LinklibrariesynamicloadPython3:Interpreter:C:/ProgramFiles/Python312/python.exe(ver3.12.4)LibrariesptimizedC:/ProgramFiles/Python312/libs/python312.libdebugC:/ProgramFiles/Python312/libs/python312_d.lib(ver3.12.4)LimitedAPI:NOnumpy:C:/ProgramFiles/Python312/Lib/site-packages/numpy/_core/include(ver2.0.0)installpath:C:/ProgramFiles/Python312/Lib/site-packages/cv2/python-3.12Python(forbuild):C:/ProgramFiles/Python312/python.exeJava:ant:NOJava:NOJNI:NOJavawrappers:NOJavatests:NOInstallto:/ProgramFiles/opencv/opencv_cp312_410/opencv_410_cp312_amd64_bulid/install-----------------------------------------------------------------Configuringdone(10.3s)Generatingdone(13.0s) 四、使用VS Studio进行编译1、点击openproject自动打开vsstudio​注意：该opencv_python3是不可或缺的。没有的话检查是否勾选BUILD_opencv_python3 2、点击选择Release。 3、右键Install，点击生成​时间大概1个小时-2个小时左右，按cpu性能而定。若没有失败，则编译成功。至此编译完成！！！五、添加环境变量，修改config.py文件1、修改config.py文件 编译完成后只有opencv/build/Install文件夹（本文命名为opencv_cp312_410\opencv_410_cp312_amd64_bulid\opencv_410_cp312_amd64 文件夹）是我们所需要的，将Install文件夹移动到你所需的位置后，需要进行如下更改：1、打开python安装路径，或者在cmake配置信息中取得，打开安装路径中的cv2文件夹（本文installpath为C:/ProgramFiles/Python312/Lib/site-packages/cv2/python-3.12，则打开C:/ProgramFiles/Python312/Lib/site-packages/cv2 ） 2、打开cv2文件夹下的config.py下的 3、编辑该路径为install的文件路径2、修改环境变量 1、打开系统变量中的Path 2、将*（install的文件路径）*/install/x64/vc17/bin添加到系统变量中 3、修改环境变量后重启计算机！！！！编译完成文件下载链接：（仅参考）百度云链接opencv_cp312_410_win64 提取码：toi4 下一篇：测试篇windows：Opencv使用Cmake&VS编译，C++、python、Cuda、Qt环境详细教学及踩坑二：测试篇-CSDN博客参考链接：AMDCustomerCommunity最新|OpenCV4.8+CUDA+扩展模块支持编译指南-腾讯云开发者社区-腾讯云https://www.cnblogs.com/guojin-blogs/p/17939955#4-%E9%A1%B9%E7%9B%AE%E6%B5%8B%E8%AF%95基于QT5.14.2的OPENCV4.8.0的cuda11.4开发环境搭建和源码编译_opencv4.8编译-CSDN博客Windows下配置CMake（入门级教程，适合新人收藏学习）-阿里云开发者社区windows下cmake的安装与环境变量配置教程_win10配置cmake环境变量-CSDN博客windows下C++openCV配置及x86编译(傻瓜式教程)_opencvx86-CSDN博客opencv源码编译及配置完整版教程（win10+vs2019+opencv-4.4.0+opencv_contrib-4.4.0）_、.4,vcvedv-CSDN博客WIN10下使用MinGW编译安装运行OpenCV_mingw编译opencv-CSDN博客https://www.cnblogs.com/liushuiruobing/p/17695059.html【opencv】【GPU】windows10下opencv4.8.0-cudaPython版本源码编译教程_cv2的版本4.8.0-CSDN博客OpenCV4.8GPU版本CMake编译详细步骤与CUDA代码演示-CSDN博客【opencv】【GPU】windows10下opencv4.8.0-cudaPython版本源码编译教程_opencv_牙牙要健康-开放原子开发者工作坊在windows上编译并使用opencv的正确姿势|Aynakeya'sBlogWindows下VisualStudio编译OpenCV详细教程（win10+vs2017+opencv4.5.2)（OpenCV最新版）（特别适合无法科学上网的童鞋们）_windows环境下vsstudio编译opencv-CSDN博客Windows编译OpenCV及扩展模块-CSDN博客