前言

  移植opencv到海思平臺，opencv支援對視訊進行解碼，需要對應的ffmpeg支援。

Ffmpeg的移植

  Ffmpeg的移植請參考之前的文章：《 FFmpeg開發筆記（十）：ffmpeg在ubuntu上的交叉編譯移植到海思HI35xx平臺》
  （PS：ffmpeg解碼h264是不需要依賴libx264的，但是編碼是需要依賴libx264的，在海思上使用了mpp的venc，所以無需libx264）

OpenCV下載

  下載地址：
  QQ群： 1047134658（點選“檔案”搜尋“ opencv”，群內與博文同步更新）

OpenCV的移植

步驟一：下載解壓

步驟二：配置cmake

export PKG_CONFIG_PATH=$PKG_CONFIG_PATH:/home/yang/work/ffmpeg_rootfs/lib/pkgconfig

  以上是目錄，注意這裡有坑（卡了一天）：

• 不要使用cmake-gui：要使用命令列，命令列可以識別到ffmpeg，但是cmake-gui不行
• 系統本身是否安裝了ffmpeg：預設是會先使用本地的，儘管設定了PKG_CONFIG_PATH
• ffmpeg否依賴x264：可檢視.pc檔案，是否有-lx264（筆者就是編譯的時候不小心編譯了libx264庫，所以cmake的時候都無法定義HAVA__FFMPEG，如果依賴的，記得將libx264拷貝到ffmpeg目錄下也行。

cmake -D CMAKE_BUILD_TYPE=RELEASE  \
             -D CMAKE_INSTALL_PREFIX=/home/yang/work/opencv_rootfs \
             -D CMAKE_C_COMPILER=aarch64-himix100-linux-gcc \
             -D CMAKE_CXX_COMPILER=aarch64-himix100-linux-g++ \
             -D BUILD_SHARED_LIBS=ON \
             -D CMAKE_CXX_FLAGS=-fPIC \
             -D CMAKE_C_FLAGS=-fPIC \
             -D CMAKE_EXE_LINKER_FLAGS=-lpthread -ldl \
             -D ENABLE_PIC=ON \
             -D WITH_1394=OFF \
             -D WITH_ARAVIS=OFF \
             -D WITH_ARITH_DEC=ON \
             -D WITH_ARITH_ENC=ON \
             -D WITH_CLP=OFF \
             -D WITH_CUBLAS=OFF \
             -D WITH_CUDA=OFF \
             -D WITH_CUFFT=OFF \
             -D WITH_FFMPEG=ON \
             -D WITH_GSTREAMER=ON \
             -D WITH_GSTREAMER_0_10=OFF \
             -D WITH_HALIDE=OFF \
             -D WITH_HPX=OFF \
             -D WITH_IMGCODEC_HDR=ON \
             -D WITH_IMGCODEC_PXM=ON \
             -D WITH_IMGCODEC_SUNRASTER=ON \
             -D WITH_INF_ENGINE=OFF \
             -D WITH_IPP=OFF \
             -D WITH_ITT=OFF \
             -D WITH_JASPER=ON \
             -D WITH_JPEG=ON \
             -D WITH_LAPACK=ON \
             -D WITH_LIBREALSENSE=OFF \
             -D WITH_NVCUVID=OFF \
             -D WITH_OPENCL=OFF \
             -D WITH_OPENCLAMDBLAS=OFF \
             -D WITH_OPENCLAMDFFT=OFF \
             -D WITH_OPENCL_SVM=OFF \
             -D WITH_OPENEXR=OFF \
             -D WITH_OPENGL=OFF \
             -D WITH_OPENMP=OFF \
             -D WITH_OPENNNI=OFF \
             -D WITH_OPENNNI2=OFF \
             -D WITH_OPENVX=OFF \
             -D WITH_PNG=OFF \
             -D WITH_PROTOBUF=OFF \
             -D WITH_PTHREADS_PF=ON \
             -D WITH_PVAPI=OFF \
             -D WITH_QT=OFF \
             -D WITH_QUIRC=OFF \
             -D WITH_TBB=OFF \
             -D WITH_TIFF=ON \
             -D WITH_VULKAN=OFF \
             -D WITH_WEBP=ON \
             -D WITH_XIMEA=OFF \
             -D BUILD_opencv_highgui=OFF \
             ..

步驟三：編譯

make -j8

步驟四：安裝

make install

部署到板子上

• 方法一：拷貝進入系統目錄，對應bin、include、lib
• 方法二：單獨一個目錄，執行前跑指令碼進入路徑（筆者是Qt，將opencv打入qt的安裝目錄，使用qt的路徑即可）。
  貼上開發本Qt的指令碼（注意系統不同，平臺的引數最後2個可能不同）：
    

Qt的引入帶ffmpeg的opencv配置檔案

# ffmpeg-3.4.8 sdk
INCLUDEPATH += /home/yang/work/ffmpeg_rootfs/include \
LIBS += -L/home/yang/work/ffmpeg_rootfs/ffmpeg-3.4.8/bin
LIBS += -lavcodec \
        -lavdevice \
        -lavfilter \
        -lavformat \
        -lavutil \
        -lswresample \
        -lswscale
INCLUDEPATH += /home/yang/work/opencv_rootfs/include \
               /home/yang/work/opencv_rootfs/include/opencv \
               /home/yang/work/opencv_rootfs/include/opencv2
LIBS += -L/home/yang/work/opencv_rootfs/lib \
        -lopencv_calib3d \
        -lopencv_core \
        -lopencv_features2d \
        -lopencv_flann \
        -lopencv_imgcodecs \
        -lopencv_imgproc \
        -lopencv_ml \
        -lopencv_objdetect \
        -lopencv_photo \
        -lopencv_shape \
        -lopencv_stitching \
        -lopencv_superres \
        -lopencv_video \
        -lopencv_videoio \
        -lopencv_videostab

板子上Qt+OpenCV播放器測試1

  將opencv部署到板子上，然後用qt的opencv播放器測試，播放慢速（判斷是解碼720p，mp4，h264大概11-12ms一幀，顯示大概15ms一幀，刷圖跟重新整理的區域面積有直接關係，所以確認海思用qt動態重新整理繪圖能力可以說得上是有點差的）（圖略）

上一篇：《 OpenCV開發筆記（七十三）：紅胖子8分鐘帶你使用opencv+dnn+yolov3識別物體》
下一篇：持續補充中…

若該文為原創文章，轉載請註明原文出處
本文章部落格地址： https://blog.csdn.net/qq21497936/article/details/123696821