在Linux下编译OpenCV_Contrib

2019-11-06 08:15:07

字体：大中小

来源：转载

供稿：网友

在linux下编译OpenCV_Contrib

最近做人脸识别项目的时候，发现在OpenCV3包中包含了人脸检测算法所需的CascadeClassifier，但不含识别算法所需的FaceRecognizer。使用官网例程时，编译器会提示缺少opencv2/contrib/contrib.hpp。查询资料后得知，人脸识别等算法模块由于不够稳定，在OpenCV3版本中，被放在了OpenCV_Contrib包里，需要单独下载并编译。

1.下载所需工具

下载opencv_contrib源码：https://github.com/Itseez/opencv_contrib下载opencv源码：https://sourceforge.net/PRojects/opencvlibrary下载安装cmake：sudo apt install cmake 顺便一提，现在从github上下载大型项目速度极慢且容易失败，建议使用墙内的coding.net导入github项目后下载。

2.配置

新建一个文件夹作为编译目录，切换到这个目录配置cmake，语法为cmake -DOPENCV_EXTRA_MODULES_PATH=<opencv_contrib>/modules <opencv_source_directory>其中opencv_contrib和opencv_source_directory分别为opencv_contrib和opencv源码解压后的文件夹如果不需要编译所有的module，可以加上-DBUILD_opencv_* 参数，例如cmake -DOPENCV_EXTRA_MODULES_PATH=<opencv_contrib>/modules -DBUILD_opencv_legacy=OFF <opencv_source_directory>。此外，如果想加快编译速度，可以跳过tests，参数为 -DBUILD_TESTS=OFF以只编译人脸识别相关库为例，配置参数为

cmake -DOPENCV_EXTRA_MODULES_PATH=<opencv_contrib>/modules -DBUILD_opencv_aruco=OFF   -DBUILD_opencv_bgsegm=OFF  -DBUILD_opencv_bioinspired=OFF  -DBUILD_opencv_ccalib=OFF  -DBUILD_opencv_cnn_3dobj=OFF  -DBUILD_opencv_contrib_world=OFF  -DBUILD_opencv_cvv=OFF  -DBUILD_opencv_datasets=OFF  -DBUILD_opencv_dnn=OFF  -DBUILD_opencv_dnns_easily_fooled=OFF  -DBUILD_opencv_dnn_modern=OFF  -DBUILD_opencv_dpm=OFF   -DBUILD_opencv_freetype=OFF  -DBUILD_opencv_fuzzy=OFF  -DBUILD_opencv_hdf=OFF  -DBUILD_opencv_line_descriptor=OFF  -DBUILD_opencv_matlab=OFF  -DBUILD_opencv_optflow=OFF  -DBUILD_opencv_phase_unwrapping=OFF  -DBUILD_opencv_plot=OFF  -DBUILD_opencv_reg=OFF  -DBUILD_opencv_rgbd=OFF  -DBUILD_opencv_saliency=OFF  -DBUILD_opencv_sfm=OFF  -DBUILD_opencv_stereo=OFF  -DBUILD_opencv_structured_light=OFF  -DBUILD_opencv_surface_matching=OFF  -DBUILD_opencv_text=OFF  -DBUILD_opencv_tracking=OFF  -DBUILD_opencv_xfeatures2d=OFF  -DBUILD_opencv_ximgproc=OFF  -DBUILD_opencv_xobjdetect=OFF  -DBUILD_opencv_xphoto=OFF -DBUILD_TESTS=OFF <opencv_source_directory>