在Linux上编译带CUDA的OpenCV

发表于 2021-02-05 更新于 2023-10-24 分类于 4-框架， A-OpenCV 阅读次数：本文字数： 14k 阅读时长 ≈ 12 分钟

本文介绍编译包含cuda模块的opencv，可以在GPU上完成opencv的操作，加速opencv的处理速度，本文用于记录在Linux上编译opencv的过程、问题

编译环境：

系统：Ubuntu 16.04
显卡：RTX3090
cmake：3.19.3
gcc：6.5.0

注：cmake、gcc系统自带版本不是3.19.3、6.5.0，为解决cmake过程中文件下载问题，升级cmake到3.19.3，但是感觉没有作用；gcc原始是5.x.x，升级之后避免了一些错误，是有效的

软件依赖：

cuda 11.1
cudnn 8.0.5

默认在以上环境下，编译opencv4.4.0，包括C++和Python3的接口。（编译v4.4.0之前，使用v4.2.0进行编译，一直无法找到cudnn，可能是cuda及cudnn版本较高，较低版本的opencv还未适配，所以建议对opencv4.4.0及以上版本进行编译，类似经历出现在^[1]）

下载源码

从github下载opencv4.4.0及opencv_contrib-4.4.0

解压后，将opencv_contrib-4.4.0放进opencv4.4.0目录下，并新建编译文件夹（build），文件目录结构如下：

安装依赖

以下参考：How to install OpenCV 4.2.0 with CUDA 10.1 on Ubuntu 20.04 LTS (Focal Fossa)

#更新系统
sudo apt update
sudo apt upgrade

#编译工具
sudo apt install build-essential cmake pkg-config unzip yasm git checkinstall

#Image I/O libs
sudo apt install libjpeg-dev libpng-dev libtiff-dev

#Video/Audio Libs — FFMPEG, GSTREAMER, x264 and so on.
sudo apt install libavcodec-dev libavformat-dev libswscale-dev libavresample-dev 
sudo apt install libgstreamer1.0-dev libgstreamer-plugins-base1.0-dev 
sudo apt install libxvidcore-dev x264 libx264-dev libfaac-dev libmp3lame-dev libtheora-dev  
sudo apt install libfaac-dev libmp3lame-dev libvorbis-dev

#OpenCore — Adaptive Multi Rate Narrow Band (AMRNB) and Wide Band (AMRWB) speech codec
sudo apt install libopencore-amrnb-dev libopencore-amrwb-dev

#Cameras programming interface libs
sudo apt-get install libdc1394-22 libdc1394-22-dev libxine2-dev libv4l-dev v4l-utils 
cd /usr/include/linux 
sudo ln -s -f ../libv4l1-videodev.h videodev.h 

#GTK lib for the graphical user functionalites coming from OpenCV highghui module
sudo apt-get install libgtk-3-dev

#Python libraries for python3（本文安装至conda下，无次步骤）
sudo apt-get install python3-dev python3-pip
sudo -H pip3 install -U pip numpy
sudo apt install python3-testresources

#Parallelism library C++ for CPU
sudo apt-get install libtbb-dev

#Optimization libraries for OpenCV
sudo apt-get install libatlas-base-dev gfortran

#Optional libraries
sudo apt-get install libprotobuf-dev protobuf-compiler 
sudo apt-get install libgoogle-glog-dev libgflags-dev 
sudo apt-get install libgphoto2-dev libeigen3-dev libhdf5-dev doxygen

cmake生成待编译文件

进入build目录下，执行以下命令

cmake -D CMAKE_BUILD_TYPE=RELEASE \
-D CMAKE_INSTALL_PREFIX=/usr/local \
-D CMAKE_C_COMPILER=/usr/bin/gcc-6 \
-D INSTALL_PYTHON_EXAMPLES=ON \
-D INSTALL_C_EXAMPLES=ON \
-D OPENCV_ENABLE_NONFREE=ON \
-D BUILD_opencv_python3=ON \
-D WITH_CUDA=ON \
-D WITH_CUDNN=ON \
-D WITH_TBB=ON \
-D OPENCV_DNN_CUDA=ON \
-D ENABLE_FAST_MATH=1 \
-D CUDA_FAST_MATH=1 \
-D CUDA_ARCH_BIN=8.6 \
-D WITH_CUBLAS=1 \
-D OPENCV_GENERATE_PKGCONFIG=ON \
-D OPENCV_EXTRA_MODULES_PATH=/home/xx/soft/opencv_gpu/opencv-4.4.0/opencv_contrib-4.4.0/modules \
-D PYTHON3_EXECUTABLE=/home/xx/anaconda3/envs/py37/bin/python3.7m \
-D PYTHON3_INCLUDE_DIR=/home/xx/anaconda3/envs/py37/include/python3.7m \
-D PYTHON3_LIBRARY=/home/xx/anaconda3/envs/py37/lib/libpython3.7m.so \
-D PYTHON3_NUMPY_INCLUDE_DIRS=/home/xx/anaconda3/envs/py37/lib/python3.7/site-packages/numpy/core/include \
-D PYTHON3_PACKAGES_PATH=/home/xx/anaconda3/envs/py37/lib/python3.7/site-packages \
-D PYTHON_DEFAULT_EXECUTABLE=/home/xx/anaconda3/envs/py37/bin/python3.7m \
-D CUDNN_LIBRARY=/usr/local/cuda/lib64/libcudnn.so.8.0.5 \
-D CUDNN_INCLUDE_DIR=/usr/local/cuda/include  \
-D CUDA_CUDA_LIBRARY=/usr/local/cuda/lib64/stubs/libcuda.so \
-D OPENCV_PYTHON3_INSTALL_PATH=/home/xx/anaconda3/envs/py37/lib/python3.7/site-packages \
-D WITH_WEBP=OFF \
-D WITH_OPENCL=OFF \
-D ETHASHLCL=OFF \
-D ENABLE_CXX11=ON \
-D BUILD_EXAMPLES=OFF \
-D OPENCV_ENABLE_NONFREE=ON \
-D WITH_OPENGL=ON \
-D WITH_GSTREAMER=ON \
-D WITH_V4L=ON \
-D WITH_QT=OFF \
-D BUILD_opencv_python3=ON \
-D BUILD_opencv_python2=OFF \
-D HAVE_opencv_python3=ON   ..

关键参数说明：

1
2
3

BUILD_opencv_python3：
CUDA_ARCH_BIN:显卡算力,Nvidia官网查询，RTX3090对应8.6
OPENCV_GENERATE_PKGCONFIG：生成pkg-config，这个务必打开，不然安装成功找不到opencv

cmake后最终确认得到以下输出：

-- General configuration for OpenCV 4.4.0 =====================================
Version control:               unknown
-- 
Extra modules:
  Location (extra):            /home/xx/soft/opencv_gpu/opencv-4.4.0/opencv_contrib-4.4.0/modules
  Version control (extra):     unknown
-- 
Platform:
  Timestamp:                   2021-02-05T02:31:19Z
  Host:                        Linux 4.15.0-133-generic x86_64
  CMake:                       3.19.3
  CMake generator:             Unix Makefiles
  CMake build tool:            /usr/bin/make
  Configuration:               RELEASE
-- 
CPU/HW features:
  Baseline:                    SSE SSE2 SSE3
    requested:                 SSE3
  Dispatched code generation:  SSE4_1 SSE4_2 FP16 AVX AVX2 AVX512_SKX
    requested:                 SSE4_1 SSE4_2 AVX FP16 AVX2 AVX512_SKX
    SSE4_1 (17 files):         + SSSE3 SSE4_1
    SSE4_2 (2 files):          + SSSE3 SSE4_1 POPCNT SSE4_2
    FP16 (1 files):            + SSSE3 SSE4_1 POPCNT SSE4_2 FP16 AVX
    AVX (5 files):             + SSSE3 SSE4_1 POPCNT SSE4_2 AVX
    AVX2 (31 files):           + SSSE3 SSE4_1 POPCNT SSE4_2 FP16 FMA3 AVX AVX2
    AVX512_SKX (7 files):      + SSSE3 SSE4_1 POPCNT SSE4_2 FP16 FMA3 AVX AVX2 AVX_512F AVX512_COMMON AVX512_SKX
-- 
C/C++:
  Built as dynamic libs?:      YES
  C++ standard:                11
  C++ Compiler:                /usr/bin/c++  (ver 6.5.0)
  C++ flags (Release):         -fsigned-char -ffast-math -W -Wall -Werror=return-type -Werror=non-virtual-dtor -Werror=address -Werror=sequence-point -Wformat -Werror=format-security -Wmissing-declarations -Wundef -Winit-self -Wpointer-arith -Wshadow -Wsign-promo -Wuninitialized -Winit-self -Wno-psabi -Wsuggest-override -Wno-delete-non-virtual-dtor -Wno-comment -fdiagnostics-show-option -Wno-long-long -pthread -fomit-frame-pointer -ffunction-sections -fdata-sections  -msse -msse2 -msse3 -fvisibility=hidden -fvisibility-inlines-hidden -O3 -DNDEBUG  -DNDEBUG
  C++ flags (Debug):           -fsigned-char -ffast-math -W -Wall -Werror=return-type -Werror=non-virtual-dtor -Werror=address -Werror=sequence-point -Wformat -Werror=format-security -Wmissing-declarations -Wundef -Winit-self -Wpointer-arith -Wshadow -Wsign-promo -Wuninitialized -Winit-self -Wno-psabi -Wsuggest-override -Wno-delete-non-virtual-dtor -Wno-comment -fdiagnostics-show-option -Wno-long-long -pthread -fomit-frame-pointer -ffunction-sections -fdata-sections  -msse -msse2 -msse3 -fvisibility=hidden -fvisibility-inlines-hidden -g  -O0 -DDEBUG -D_DEBUG
  C Compiler:                  /usr/bin/gcc-6
  C flags (Release):           -fsigned-char -ffast-math -W -Wall -Werror=return-type -Werror=non-virtual-dtor -Werror=address -Werror=sequence-point -Wformat -Werror=format-security -Wmissing-declarations -Wmissing-prototypes -Wstrict-prototypes -Wundef -Winit-self -Wpointer-arith -Wshadow -Wuninitialized -Winit-self -Wno-psabi -Wno-comment -fdiagnostics-show-option -Wno-long-long -pthread -fomit-frame-pointer -ffunction-sections -fdata-sections  -msse -msse2 -msse3 -fvisibility=hidden -O3 -DNDEBUG  -DNDEBUG
  C flags (Debug):             -fsigned-char -ffast-math -W -Wall -Werror=return-type -Werror=non-virtual-dtor -Werror=address -Werror=sequence-point -Wformat -Werror=format-security -Wmissing-declarations -Wmissing-prototypes -Wstrict-prototypes -Wundef -Winit-self -Wpointer-arith -Wshadow -Wuninitialized -Winit-self -Wno-psabi -Wno-comment -fdiagnostics-show-option -Wno-long-long -pthread -fomit-frame-pointer -ffunction-sections -fdata-sections  -msse -msse2 -msse3 -fvisibility=hidden -g  -O0 -DDEBUG -D_DEBUG
  Linker flags (Release):      -Wl,--gc-sections -Wl,--as-needed  
  Linker flags (Debug):        -Wl,--gc-sections -Wl,--as-needed  
  ccache:                      NO
  Precompiled headers:         NO
  Extra dependencies:          m pthread cudart_static dl rt nppc nppial nppicc nppidei nppif nppig nppim nppist nppisu nppitc npps cublas cudnn cufft -L/usr/local/cuda/lib64 -L/usr/lib/x86_64-linux-gnu
  3rdparty dependencies:
-- 
OpenCV modules:
  To be built:                 alphamat aruco bgsegm bioinspired calib3d ccalib core cudaarithm cudabgsegm cudacodec cudafeatures2d cudafilters cudaimgproc cudalegacy cudaobjdetect cudaoptflow cudastereo cudawarping cudev datasets dnn dnn_objdetect dnn_superres dpm face features2d flann freetype fuzzy gapi hdf hfs highgui img_hash imgcodecs imgproc intensity_transform line_descriptor ml objdetect optflow phase_unwrapping photo plot python3 quality rapid reg rgbd saliency sfm shape stereo stitching structured_light superres surface_matching text tracking ts video videoio videostab xfeatures2d ximgproc xobjdetect xphoto
  Disabled:                    python2 world
  Disabled by dependency:      -
  Unavailable:                 cnn_3dobj cvv java js julia matlab ovis viz
  Applications:                tests perf_tests apps
  Documentation:               NO
  Non-free algorithms:         YES
-- 
GUI: 
  GTK+:                        YES (ver 3.18.9)
    GThread :                  YES (ver 2.48.2)
    GtkGlExt:                  NO
  OpenGL support:              NO
  VTK support:                 NO
-- 
Media I/O: 
  ZLib:                        /usr/lib/x86_64-linux-gnu/libz.so (ver 1.2.8)
  JPEG:                        /usr/lib/x86_64-linux-gnu/libjpeg.so (ver 80)
  PNG:                         /usr/lib/x86_64-linux-gnu/libpng.so (ver 1.2.54)
  TIFF:                        /usr/lib/x86_64-linux-gnu/libtiff.so (ver 42 / 4.0.6)
  JPEG 2000:                   OpenJPEG (ver 2.4.0)
  OpenEXR:                     build (ver 2.3.0)
  HDR:                         YES
  SUNRASTER:                   YES
  PXM:                         YES
  PFM:                         YES
-- 
Video I/O:
  DC1394:                      YES (2.2.4)
  FFMPEG:                      YES
    avcodec:                   YES (56.60.100)
    avformat:                  YES (56.40.101)
    avutil:                    YES (54.31.100)
    swscale:                   YES (3.1.101)
    avresample:                YES (2.1.0)
  GStreamer:                   YES (1.8.3)
  v4l/v4l2:                    YES (linux/videodev2.h)
-- 
Parallel framework:            TBB (ver 4.4 interface 9002)
-- 
Trace:                         YES (with Intel ITT)
-- 
Other third-party libraries:
  Lapack:                      YES (/usr/lib/libopenblas.so)
  Eigen:                       YES (ver 3.2.92)
  Custom HAL:                  NO
  Protobuf:                    build (3.5.1)
-- 
NVIDIA CUDA:                   YES (ver 11.1, CUFFT CUBLAS FAST_MATH)
  NVIDIA GPU arch:             86
  NVIDIA PTX archs:
-- 
cuDNN:                         YES (ver 8.0.5)
-- 
Python 3:
  Interpreter:                 /home/xx/anaconda3/envs/py37/bin/python3.7m (ver 3.7.9)
  Libraries:                   /home/xx/anaconda3/envs/py37/lib/libpython3.7m.so (ver 3.7.9)
  numpy:                       /home/xx/anaconda3/envs/py37/lib/python3.7/site-packages/numpy/core/include (ver 1.19.2)
  install path:                /home/xx/anaconda3/envs/py37/lib/python3.7/site-packages/cv2/python-3.7
-- 
Python (for build):            /home/xx/anaconda3/envs/py37/bin/python3.7m
-- 
Java:                          
  ant:                         NO
  JNI:                         NO
  Java wrappers:               NO
  Java tests:                  NO
-- 
Install to:                    /usr/local

注：系统用户名使用xx替代

gcc编译

在build目录下执行以下命令

make -j16

注：虽然直接make出问题的概率更小，但是为了加速编译速度，这里指使用16个多线程进行编译，这里多线程数量一般不大于CPU核心数，可以通过nproc命令查看CPU核心数。^[2]

安装到系统

在build目录下执行以下命令

1	sudo make install

配置环境变量

执行以下命令

1 2	sudo /bin/bash -c 'echo "/usr/local/lib" >> /etc/ld.so.conf.d/opencv.conf' sudo ldconfig

如果想将opencv安装到Python，建议在cmake编译参数中使用-D OPENCV_PYTHON3_INSTALL_PATH指定安装目录

1	-D OPENCV_PYTHON3_INSTALL_PATH=/home/xx/anaconda3/envs/py37/lib/python3.7/site-packages \

查看opencv是否安装成功

1 2	pkg-config --modversion opencv pkg-config --libs opencv4

编译过程遇到的问题

编译opencv主要有两个过程，cmake和make，错误主要出现在cmake阶段，根据出现的错误情况，主要分为以下几类

下载失败

cmake过程中，需要下载的文件会放在opencv4.4.0下的隐藏目录.cache下，如果下载失败，可以手动下载，放入相应目录解决。

xfeatures2d^[3]

boostdesc_bgm.i
boostdesc_bgm_bi.i
boostdesc_bgm_hd.i
boostdesc_lbgm.i
boostdesc_binboost_064.i
boostdesc_binboost_128.i
boostdesc_binboost_256.i
vgg_generated_120.i
vgg_generated_64.i
vgg_generated_80.i
vgg_generated_48.i

进入opencv4.4.0下的隐藏目录.cache/xfeatures2d，执行以下命令解决：

cd boostdesc
curl https://raw.githubusercontent.com/opencv/opencv_3rdparty/34e4206aef44d50e6bbcd0ab06354b52e7466d26/boostdesc_lbgm.i > 0ae0675534aa318d9668f2a179c2a052-boostdesc_lbgm.i
curl https://raw.githubusercontent.com/opencv/opencv_3rdparty/34e4206aef44d50e6bbcd0ab06354b52e7466d26/boostdesc_binboost_256.i > e6dcfa9f647779eb1ce446a8d759b6ea-boostdesc_binboost_256.i
curl https://raw.githubusercontent.com/opencv/opencv_3rdparty/34e4206aef44d50e6bbcd0ab06354b52e7466d26/boostdesc_binboost_128.i > 98ea99d399965c03d555cef3ea502a0b-boostdesc_binboost_128.i
curl https://raw.githubusercontent.com/opencv/opencv_3rdparty/34e4206aef44d50e6bbcd0ab06354b52e7466d26/boostdesc_binboost_064.i > 202e1b3e9fec871b04da31f7f016679f-boostdesc_binboost_064.i
curl https://raw.githubusercontent.com/opencv/opencv_3rdparty/34e4206aef44d50e6bbcd0ab06354b52e7466d26/boostdesc_bgm_hd.i > 324426a24fa56ad9c5b8e3e0b3e5303e-boostdesc_bgm_hd.i
curl https://raw.githubusercontent.com/opencv/opencv_3rdparty/34e4206aef44d50e6bbcd0ab06354b52e7466d26/boostdesc_bgm_bi.i > 232c966b13651bd0e46a1497b0852191-boostdesc_bgm_bi.i
curl https://raw.githubusercontent.com/opencv/opencv_3rdparty/34e4206aef44d50e6bbcd0ab06354b52e7466d26/boostdesc_bgm.i > 0ea90e7a8f3f7876d450e4149c97c74f-boostdesc_bgm.i
cd vgg
curl https://raw.githubusercontent.com/opencv/opencv_3rdparty/fccf7cd6a4b12079f73bbfb21745f9babcd4eb1d/vgg_generated_120.i > 151805e03568c9f490a5e3a872777b75-vgg_generated_120.i
curl https://raw.githubusercontent.com/opencv/opencv_3rdparty/fccf7cd6a4b12079f73bbfb21745f9babcd4eb1d/vgg_generated_64.i > 7126a5d9a8884ebca5aea5d63d677225-vgg_generated_64.i
curl https://raw.githubusercontent.com/opencv/opencv_3rdparty/fccf7cd6a4b12079f73bbfb21745f9babcd4eb1d/vgg_generated_48.i > e8d0dcd54d1bcfdc29203d011a797179-vgg_generated_48.i
curl https://raw.githubusercontent.com/opencv/opencv_3rdparty/fccf7cd6a4b12079f73bbfb21745f9babcd4eb1d/vgg_generated_80.i > 7cd47228edec52b6d82f46511af325c5-vgg_generated_80.i

注：如何curl无法下载，可以通过浏览器挨个下载，并按照以上进行命名各文件

ippicv

手动从github下载，放入.cache/ppicv目录下

**face_landmark_model.dat **

手动从github下载，放入.cache/data目录下，注意文件名前部是该文件的md5值，可以通过命令md5sum file计算该值

软件未安装（Not Found）

此类错误比较简单，缺什么安装什么，比如安装过程中遇到以下缺失软件的解决办法

tesserocr安装失败^[4]

1 2	sudo apt-get install libleptonica-dev libtesseract-dev python -m pip install tesserocr

lapacke.h缺失^[5]

明明已经按照该软件，但是OpenBLAS一致没找到该文件，但是搜索文件发现该文件在/usr/include/下，只能手动拷贝ls /usr/include/lapacke*文件至/usr/include/openblas/目录下

Could NOT find CUDNN: Found unsuitable version “…”, but required is at least “7.5” (found /usr/local/cuda-10.2/lib64/libcudnn.so)

这是在编译opencv4.2.0时出现的错误，实际已经安装cudnn，也满足7.5以上的要求，但是就是找不，可以在cmake中加入参数：-D CUDNN_VERSION='8.0'解决^[1:1]，但是后续还会遇到其他问题，编译opencv4.4.0时，该问题不再出现

No package ‘gtk±3.0’ found

1	sudo apt-get install libgtk-3-dev

卸载opencv

通过源码安装的opencv，可以进入编译目录下（build）执行以下命令，卸载opencv

1	sudo make uninstall

使用cuda模块的简单例子

查看opencv的cuda模块支持的功能，进入python终端，输入以下命令

1
2
3

import cv2
dir(cv2.cuda)
dir(cv2.cuda_GpuMat())

例子：在GPU做resize

#读取图片
frame=cv2.imread('test.jpg')

#上传到gpu进行处理
gpu_frame=cv2.cuda_GpuMat()
gpu_frame.upload(frame)
print(gpu_frame.cudaPtr())

#resize
gpu_resframe=cv2.cuda.resize(gpu_frame,(1024,512))
cpu_resfram=gpu_resframe.download()
print(cpu_resfram.shape)