ubuntu16.04下opencv3.3 GPU（CUDA）加速

在opencv的cuda加速版时候会将cuda的一个libcudart.so.8.0加入到opencv的lib中，所以如果你以后升级你的cuda，那么opencv需要卸载后重新编译才能使用，毕竟cuda变了，原来的libcudart.so.8.0也不能继续调用了。

1.准备

1.安装好ubuntu16.04
2.安装配置好opencv3.3.0 配置教程
3.安装好cuda8.0
4.安装QT creator(可选)

2.编译

opencv的GPU加速需要用到cuda,故需要在此之前的基础上重新编译一次

cd opencv-3.3.0   
mkdir my_build_dir_gpu   
cd my_build_dir_gpu    

 

cmake   -D CMAKE_BUILD_TYPE=RELEASE   -D CMAKE_INSTALL_PREFIX=/usr/local   -D WITH_TBB=ON   -D WITH_V4L=ON   -D WITH_QT=ON   -D WITH_OPENGL=ON   -D WITH_CUDA=ON   -D ENABLE_FAST_MATH=1   -D CUDA_FAST_MATH=1   -D CUDA_NVCC_FLAGS="-D_FORCE_INLINES"   -D WITH_CUBLAS=1 \..

 

当出现以上现象时表示已经cmake成功

接下来就是漫长的编译过程(我编译花了3个小时 - -)，视CPU性能（我i7-6700hq - -）

make
sudo make install

3.配置.bashrc

    echo '/usr/local/lib' | sudo tee -a /etc/ld.so.conf.d/opencv.conf    
    sudo ldconfig    
    printf '# OpenCV\nPKG_CONFIG_PATH=$PKG_CONFIG_PATH:/usr/local/lib/pkgconfig\nexport PKG_CONFIG_PATH\n' >> ~/.bashrc    
    source ~/.bashrc    

4.代码测试

首先新建文件夹test，然后在test中新建文件CMakeLists.txt和main.cpp（文件内的内容看下面的代码），然后cd到test目录下，假设文件夹test的目录是/home/test，那么就使用命令

cd /home/test

接着使用命令

cmake .

如果没有cmake，请自行百度安装。这是会出现好几个文件，当中会有一个makefile，接着使用命令

make

就会有一个名为"opencv_example"的可执行文件，最后使用命令

./opencv_example

如果opencv安装成功，代码编译成功：
则会显示出1来（如果2块显卡就显示2），若为0表示安装不成功

CMakeLists.txt内容为:

# cmake needs this line
cmake_minimum_required(VERSION 2.8)

# Define project name
project(test)
# Find OpenCV, you may need to set OpenCV_DIR variable
# to the absolute path to the directory containing OpenCVConfig.cmake file
# via the command line or GUI
find_package(OpenCV 3.3.0 REQUIRED)

# If the package has been found, several variables will
# be set, you can find the full list with descriptions
# in the OpenCVConfig.cmake file.
# Print some message showing some of them
message(STATUS "OpenCV library status:")
message(STATUS "    version: ${OpenCV_VERSION}")
message(STATUS "    libraries: ${OpenCV_LIBS}")
message(STATUS "    include path: ${OpenCV_INCLUDE_DIRS}")

if(CMAKE_VERSION VERSION_LESS "2.8.11")
  # Add OpenCV headers location to your include paths
  include_directories(${OpenCV_INCLUDE_DIRS})
endif()

# Declare the executable target built from your sources
add_executable(opencv_example main.cpp)

# Link your application with OpenCV libraries
target_link_libraries(opencv_example ${OpenCV_LIBS})

---

main.cpp代码为：

using namespace std;
#include "opencv2/opencv.hpp"
#include "opencv2/gpu/gpu.hpp"
using namespace cv;

int main()
{
    int num_devices = cv::cuda::getCudaEnabledDeviceCount();

    cout<<num_devices<<endl;
}

 

---

另外一段测试代码，用来调用摄像头，然后查看是否占用gpu的，如果上面成功了，这段代码就没必要测试。

using namespace std;
#include "opencv2/opencv.hpp"
#include "opencv2/gpu/gpu.hpp"
using namespace cv;

int main()
{
    int num_devices = cv::cuda::getCudaEnabledDeviceCount();

    if(num_devices <= 0)
    {
        std::cerr<<"There is no device."<<std::endl;
        return -1;
    }
    int enable_device_id = -1;
    for(int i=0;i<num_devices;i++)
    {
        cv::cuda::DeviceInfo dev_info(i);
        if(dev_info.isCompatible())
        {
            enable_device_id=i;
        }
    }
    if(enable_device_id < 0)
    {
        std::cerr<<"GPU module isn't built for GPU"<<std::endl;
        return -1;
    }
    cv::cuda::setDevice(enable_device_id);

    std::cout<<"GPU is ready, device ID is "<<num_devices<<"\n";

    VideoCapture cap(0);
    Mat frame;
    Mat dst_image;
    while(1)
    {
        cap>>frame;
        cuda::GpuMat d_src_img(frame);
        cuda::GpuMat d_dst_img;
        cuda::cvtColor(d_src_img,d_dst_img,CV_BGR2GRAY);
        d_dst_img.download(dst_image);
        imshow("test",dst_image);
        waitKey(1);
    }
    return 0;
}

代码效果：
会显示出灰度的摄像头图像

这个时候我们可以调用

nvidia-smi

可以看到该程序使用了GPU资源

5.可能出现的问题

1.报错

/usr/include/opencv2/gpu/gpu.hpp:432: error: 'vector' does not name a type
 CV_EXPORTS void merge(const vector<GpuMat>& src, GpuMat& dst, Stream& stream = Stream::Null());

解决办法：把 using namespace std放到最最上面即可。

---

2.提示未定义gpu::getCudaEnabledDeviceCount()

/home/xukeqin/code/cpp/gpu_test/main.cpp:-1: error: undefined reference to `cv::gpu::getCudaEnabledDeviceCount()'

解决办法：opencv3.2.0中的gpu模块有比较大的变化,原来的gpu换成了现在的cuda,故改为cuda::getCudaEnabledDeviceCount()即可！

转自：https://blog.csdn.net/u013066730/article/details/53782015