目录

• 1. 前言
• 2. OPENCV交叉编译
• 3. NCNN交叉编译
• 4. OPENCV和NCNN移植

1. 前言

介绍了OpenCV和NCNN库的交叉编译过程，并在TIAM62开发板上部署YOLACT实力分割模型

2. OpenCV交叉编译

首先下载opencv源码，如下

git clone https://github.com/opencv/opencv.git

cd opencv

git submodule update --init --recursive

下载opencv_contrib源码，如下

git clone https://github.com/opencv/opencv_contrib

开启交叉编译docker环境，将opencv源码和opencv_contrib源码复制到docker中，如下

cd docker/

docker load -i myir-env.tar

docker images



// REPOSITORY   TAG       IMAGE ID       CREATED       SIZE

// myir-env     ros2      41f916f6b5c0   12 days ago   17.9GB



docker run -itd 41f916f6b5c0



docker ps 

// CONTAINER ID   IMAGE          COMMAND       CREATED             STATUS             PORTS     NAMES

// 88c3cc603221   41f916f6b5c0   "/bin/bash"   About an hour ago   Up About an hour             mystifying_dewdney



docker cp opencv 88c3cc603221:/opt/arago-2023.04/sysroots/aarch64-oe-linux/home/root



docker cp opencv_contrib 88c3cc603221:/opt/arago-2023.04/sysroots/aarch64-oe-linux/home/root/

进入docker中编译opencv源码，首先运行如下命令将编译器改为交叉编译器，

docker exec -it -u root 88c3cc603221 bash

. /opt/arago-2023.04/environment-setup-aarch64-oe-linux

然后编译opencv和opencv_contrib源码，

cd /opt/arago-2023.04/sysroots/aarch64-oe-linux/home/root/opencv

mkdir build & cd build



cmake .. -DCMAKE_INSTALL_PREFIX=/opt/arago-2023.04/sysroots/aarch64-oe-linux/usr/local/opencv/ -DOPENCV_EXTRA_MODULE_PATH=../../opencv_contrib-4.x/modules/ -DBUILD_opencv_world=OFF -DBUILD_PNG=ON -DBUILD_ZLIB=ON -DWITH_PNG=ON -DWITH_JPEG=ON -DWITH_OPENJPEG=ON -DBUILD_opencv_apps=ON -DBUILD_opencv_calib3d=ON -DBUILD_opencv_core=ON -DBUILD_opencv_dnn=ON -DBUILD_opencv_features2d=ON -DBUILD_opencv_flann=ON -DBUILD_opencv_highgui=ON -DBUILD_opencv_imgcodecs=ON -DBUILD_opencv_imgproc=ON -DBUILD_opencv_video=ON -DBUILD_opencv_videoio=ON -DWITH_OPENGL=ON -DWITH_QT=ON



make & make install

3. NCNN交叉编译

下载ncnn源码，如下

git clone https://github.com/Tencent/ncnn.git

将ncnn源码复制到docker中，

docker cp ncnn 88c3cc603221:/opt/arago-2023.04/sysroots/aarch64-oe-linux/home/root

进入docker编译ncnn源码，如下

cd /opt/arago-2023.04/sysroots/aarch64-oe-linux/home/root/ncnn

mkdir build & cd build



cmake -DCMAKE_BUILD_TYPE=Release -DNCNN_VULKAN=ON -DNCNN_BUILD_EXAMPLES=ON -DCMAKE_INSTALL_PREFIX=/opt/arago-2023.04/sysroots/aarch64-oe-linux/usr/local/ncnn/ ..



make & make install

4. opencv和ncnn移植

在/opt/arago-2023.04/sysroots/aarch64-oe-linux/usr/local/opencv/lib/和/opt/arago-2023.04/sysroots/aarch64-oe-linux/usr/local/ncnn/lib/有opencv和ncnn的库文件，将其移植到TIAM62开发板上