26、ubuntu环境下编译OPENCV的NVCODEC SDK版本进行硬件解码视频和播放测试

基本思想:编译opencv包含cuda版本,然后进行解码播放显示帧

一 、 首先本机配置了cuda和cudnn的环境，参考：https://sxj731533730.blog.csdn.net/article/details/115064903然后下载解码组件进行重新编译opencv进行准备

ubuntu@ubuntu:~/Downloads$  wget https://developer.nvidia.com/downloads/designworks/video-codec-sdk/secure/12.1/video_codec_sdk_12.1.14.zip
然后解压之后
ubuntu@ubuntu:~/Downloads$ sudo cp Video_Codec_SDK_12.1.14/Interface/* /usr/local/cuda/include

然后下载opencv进行编译我用的最新版本4.6.0 (2023-5-25)

ubuntu@ubuntu:~$ git clone https://github.com/opencv/opencv.git
ubuntu@ubuntu:~$ git clone https://github.com/opencv/opencv_contrib.git
ubuntu@ubuntu:~$ cmake -D CMAKE_BUILD_TYPE=RELEASE \
      -D CMAKE_INSTALL_PREFIX=/usr/local  \
      -D WITH_TBB=ON \
      -D BUILD_TBB=ON  \
      -D ENABLE_FAST_MATH=1 \
      -D CUDA_FAST_MATH=1 \
      -D WITH_CUBLAS=1 \
      -D WITH_V4L=ON \
      -D WITH_LIBV4L=ON \
      -D WITH_CUDA=ON \
      -D WITH_CUDNN=ON \
      -D WITH_CUDEV=ON \
      -D WITH_GTK_2_X=ON \
      -D WITH_NVCUVID=ON \
      -D CUDA_ARCH_BIN=8.6 \
      -D OPENCV_EXTRA_MODULES_PATH=../../opencv_contrib/modules \
      -D WITH_QT=ON \
      -D WITH_OPENGL=ON \
      -D WITH_FFMPEG=ON \
      -D OPENCV_GENERATE_PKGCONFIG=ON \
      ..
ubuntu@ubuntu:~/opencv/build$ sudo make install
ubuntu@ubuntu:~/opencv/build$ sudo ldconfig
#直接跳到第二步进行测试一下，要是没问题，下面这些不必操作
ubuntu@ubuntu:~/opencv/build$ cd /usr/local/lib/pkgconfig
ubuntu@ubuntu:~/opencv/build$ sudo mkdir pkgconfig && cd pkgconfig
ubuntu@ubuntu:~/opencv/build$ sudo gedit opencv.pc
----------
# Package Information for pkg-config

prefix=/usr/local
exec_prefix=${prefix}
libdir=${exec_prefix}/lib
includedir=${prefix}/include/opencv4

Name: OpenCV
Description: Open Source Computer Vision Library
Version: 4.6.0
Libs: -L${exec_prefix}/lib -lopencv_gapi -lopencv_ml -lopencv_video -lopencv_highgui -lopencv_videoio -lopencv_imgcodecs -lopencv_objdetect -lopencv_dnn -lopencv_stitching -lopencv_calib3d -lopencv_features2d -lopencv_flann -lopencv_photo -lopencv_imgproc -lopencv_core
Libs.private: -ldl -lm -lpthread -lrt
Cflags: -I${includedir}
----------
ubuntu@ubuntu:~/opencv/build$ sudo ln -s opencv4.pc opencv.pc 
ubuntu@ubuntu:~/opencv/build$ pkg-config --libs --cflags opencv4

二、测试安装

ubuntu@ubuntu:~/opencv/build$ pkg-config --libs --cflags opencv4
-I/usr/local/include/opencv4 -L/usr/local/lib -lopencv_gapi -lopencv_stitching -lopencv_alphamat -lopencv_aruco -lopencv_barcode -lopencv_bgsegm -lopencv_bioinspired -lopencv_ccalib -lopencv_cudabgsegm -lopencv_cudafeatures2d -lopencv_cudaobjdetect -lopencv_cudastereo -lopencv_cvv -lopencv_dnn_objdetect -lopencv_dnn_superres -lopencv_dpm -lopencv_face -lopencv_freetype -lopencv_fuzzy -lopencv_hdf -lopencv_hfs -lopencv_img_hash -lopencv_intensity_transform -lopencv_line_descriptor -lopencv_mcc -lopencv_quality -lopencv_rapid -lopencv_reg -lopencv_rgbd -lopencv_saliency -lopencv_stereo -lopencv_structured_light -lopencv_phase_unwrapping -lopencv_superres -lopencv_cudacodec -lopencv_surface_matching -lopencv_tracking -lopencv_highgui -lopencv_datasets -lopencv_text -lopencv_plot -lopencv_videostab -lopencv_cudaoptflow -lopencv_optflow -lopencv_cudalegacy -lopencv_videoio -lopencv_cudawarping -lopencv_viz -lopencv_wechat_qrcode -lopencv_xfeatures2d -lopencv_shape -lopencv_ml -lopencv_ximgproc -lopencv_video -lopencv_xobjdetect -lopencv_objdetect -lopencv_calib3d -lopencv_imgcodecs -lopencv_features2d -lopencv_dnn -lopencv_flann -lopencv_xphoto -lopencv_photo -lopencv_cudaimgproc -lopencv_cudafilters -lopencv_imgproc -lopencv_cudaarithm -lopencv_core -lopencv_cudev

三、测试单帧速度

cmakelists.txt

cmake_minimum_required(VERSION 3.16)
project(untitled8)
set(CMAKE_CXX_FLAGS "-std=c++11")
#message(STATUS ${OpenCV_INCLUDE_DIRS})
#添加头文件
find_package (OpenCV 4.0.0 REQUIRED)

include_directories(${OpenCV_INCLUDE_DIRS})
#链接Opencv库
add_executable(untitled8 main.cpp)
target_link_libraries(untitled8 ${OpenCV_LIBS} )

main.cpp

#include <opencv2/core.hpp>
#include <opencv2/highgui.hpp>
#include <opencv2/cudacodec.hpp>
#include <chrono>
#include "stdio.h"
#include "string"
using namespace cv;

int gpu_test()
{
    cv::cuda::printCudaDeviceInfo(cv::cuda::getDevice());
    int count = cv::cuda::getCudaEnabledDeviceCount();
    printf("GPU Device Count : %d \n", count);

     const std::string filename = "/home/ubuntu/untitled8/mot17_demo.mp4";
    cv::Ptr<cv::cudacodec::VideoReader> reader = cv::cudacodec::createVideoReader(filename);

    cv::cuda::GpuMat gpu_frame;

    int frame_id = 0;
    std::chrono::steady_clock::time_point start = std::chrono::steady_clock::now();
    while (reader->nextFrame(gpu_frame))
    {
        frame_id = frame_id + 1;

    }
    std::chrono::steady_clock::time_point end = std::chrono::steady_clock::now();
    std::chrono::duration<double> time_useds = std::chrono::duration_cast<std::chrono::duration<double>>(end - start);
    double time_ms = time_useds.count() * 1000.0f;
    double fps = double(frame_id) / time_ms * 1000.0f;
    printf("GPU test took time: %f ms, frames: %d , FPS: %f\n", time_ms, frame_id, fps);

    reader.release();

    return 0;
}

int cpu_test()
{
     const std::string filename = "/home/ubuntu/untitled8/mot17_demo.mp4";
    cv::VideoCapture capture;
    capture.open(filename);
    if (!capture.isOpened())
    {
        printf("Open video failed !!! \n");
        return -1;
    }

    int width = (int)capture.get(cv::CAP_PROP_FRAME_WIDTH);
    int height = (int)capture.get(cv::CAP_PROP_FRAME_HEIGHT);
    printf("src video width: %d , %d \n", width, height);

    cv::Mat frame;

    int frame_id = 0;
    std::chrono::steady_clock::time_point start = std::chrono::steady_clock::now();
    while (capture.read(frame))
    {
        frame_id = frame_id + 1;

    }
    std::chrono::steady_clock::time_point end = std::chrono::steady_clock::now();
    std::chrono::duration<double> time_useds = std::chrono::duration_cast<std::chrono::duration<double>>(end - start);
    double time_ms = time_useds.count() * 1000.0f;
    double fps = double(frame_id) / time_ms * 1000.0f;
    printf("CPU test took time: %f ms, frames: %d , FPS: %f\n", time_ms, frame_id, fps);

    capture.release();

    return 0;
}

int main(int argc, char const *argv[])
{

    gpu_test();

    cpu_test();

    return 0;
}

测试数据

/home/ubuntu/untitled8/cmake-build-debug/untitled8
*** CUDA Device Query (Runtime API) version (CUDART static linking) *** 

Device count: 1

Device 0: "NVIDIA GeForce RTX 3050 Laptop GPU"
  CUDA Driver Version / Runtime Version          12.0 / 11.10
  CUDA Capability Major/Minor version number:    8.6
  Total amount of global memory:                 3902 MBytes (4091478016 bytes)
  GPU Clock Speed:                               1.50 GHz
  Max Texture Dimension Size (x,y,z)             1D=(131072), 2D=(131072,65536), 3D=(16384,16384,16384)
  Max Layered Texture Size (dim) x layers        1D=(32768) x 2048, 2D=(32768,32768) x 2048
  Total amount of constant memory:               65536 bytes
  Total amount of shared memory per block:       49152 bytes
  Total number of registers available per block: 65536
  Warp size:                                     32
  Maximum number of threads per block:           1024
  Maximum sizes of each dimension of a block:    1024 x 1024 x 64
  Maximum sizes of each dimension of a grid:     2147483647 x 65535 x 65535
  Maximum memory pitch:                          2147483647 bytes
  Texture alignment:                             512 bytes
  Concurrent copy and execution:                 Yes with 2 copy engine(s)
  Run time limit on kernels:                     Yes
  Integrated GPU sharing Host Memory:            No
  Support host page-locked memory mapping:       Yes
  Concurrent kernel execution:                   Yes
  Alignment requirement for Surfaces:            Yes
  Device has ECC support enabled:                No
  Device is using TCC driver mode:               No
  Device supports Unified Addressing (UVA):      Yes
  Device PCI Bus ID / PCI location ID:           1 / 0
  Compute Mode:
      Default (multiple host threads can use ::cudaSetDevice() with device simultaneously) 

deviceQuery, CUDA Driver = CUDART, CUDA Driver Version  = 12.0, CUDA Runtime Version = 11.10, NumDevs = 1

GPU Device Count : 1 
GPU test took time: 329.191769 ms, frames: 200 , FPS: 607.548605
src video width: 1920 , 1080 
CPU test took time: 375.254235 ms, frames: 200 , FPS: 532.972000

Process finished with exit code 0

四、测试视频播放

#include <opencv2/core.hpp>
#include <opencv2/highgui.hpp>
#include <opencv2/cudacodec.hpp>
#include <opencv2/imgproc/imgproc.hpp>
#include <chrono>
#include <iomanip>
#include <opencv2/highgui/highgui.hpp>
#include "string"
using namespace cv;

int gpu_test()
{
    cv::cuda::printCudaDeviceInfo(cv::cuda::getDevice());
    int count = cv::cuda::getCudaEnabledDeviceCount();
    printf("GPU Device Count : %d \n", count);

     const std::string filename = "/home/ubuntu/untitled8/palace.mp4";
    cv::Ptr<cv::cudacodec::VideoReader> reader = cv::cudacodec::createVideoReader(filename);

    cv::cuda::GpuMat gpu_frame;
    cv::Mat frame;
    int frame_id = 0;
    auto startTime = std::chrono::steady_clock::now();
    int counter = 0;
    float fps = 0;
    while (reader->nextFrame(gpu_frame))
    {
        counter++;
        auto currentTime = std::chrono::steady_clock::now();
        auto elapsed = std::chrono::duration_cast<std::chrono::duration<float>>(currentTime - startTime);
        if (elapsed > std::chrono::seconds(1)) {
            fps = counter / elapsed.count();
            counter = 0;
            startTime = currentTime;
        }

        gpu_frame.step = gpu_frame.cols * gpu_frame.channels();
        gpu_frame.download(frame);
        std::stringstream fpsStr;
        fpsStr << "fps: " << std::fixed << std::setprecision(2) << fps;
        cv::putText(frame, fpsStr.str(), cv::Point(2,  20), cv::FONT_HERSHEY_TRIPLEX, 0.4, cv::Scalar(0,255,0));

        cv::imshow("frame",frame);
        cv::waitKey(1);
    }

    reader.release();

    return 0;
}

int cpu_test()
{
     const std::string filename = "/home/ubuntu/untitled8/palace.mp4";
    cv::VideoCapture capture;
    capture.open(filename);
    if (!capture.isOpened())
    {
        printf("Open video failed !!! \n");
        return -1;
    }

    int width = (int)capture.get(cv::CAP_PROP_FRAME_WIDTH);
    int height = (int)capture.get(cv::CAP_PROP_FRAME_HEIGHT);
    printf("src video width: %d , %d \n", width, height);

    cv::Mat frame;


    auto startTime = std::chrono::steady_clock::now();
    int counter = 0;
    float fps = 0;
    while (capture.read(frame))
    {

        counter++;
        auto currentTime = std::chrono::steady_clock::now();
        auto elapsed = std::chrono::duration_cast<std::chrono::duration<float>>(currentTime - startTime);
        if (elapsed > std::chrono::seconds(1)) {
            fps = counter / elapsed.count();
            counter = 0;
            startTime = currentTime;
        }
        std::stringstream fpsStr;
        fpsStr << "fps: " << std::fixed << std::setprecision(2) << fps;
        cv::putText(frame, fpsStr.str(), cv::Point(2,  20), cv::FONT_HERSHEY_TRIPLEX, 0.4, cv::Scalar(0,255,0));

        cv::imshow("frame",frame);
        cv::waitKey(1);


    }

    capture.release();

    return 0;
}

int main(int argc, char const *argv[])
{

   // gpu_test();

   cpu_test();

    return 0;
}

先研究opencv+cuda的程序开发

附录简单灰度图

#include <opencv2/core.hpp>
#include <opencv2/highgui.hpp>
#include <opencv2/cudacodec.hpp>
#include <chrono>
#include <opencv2/imgproc.hpp>
#include "stdio.h"
#include "opencv2/core/cuda.hpp"
#include "opencv2/cudaimgproc.hpp"

#include "string"
using namespace cv;
using namespace cv::cuda;

int main(int argc, char const *argv[])
{
    cuda::printCudaDeviceInfo(cuda::getDevice());
    int count = cuda::getCudaEnabledDeviceCount();
    printf("GPU Device Count : %d n", count);
    // 灰度转换
    Mat src_host = imread("../0.jpg");
    GpuMat src, gray;
    src.upload(src_host);
    cuda::cvtColor(src, gray, cv::COLOR_BGR2GRAY);
    Mat gray_host;
    gray.download(gray_host);
    imshow("src", src_host);
    imshow("gray", gray_host);
    waitKey(0);

    return 0;
}