CentOS6-Linux下安装OpenCV

CentOS6 - Linux下安装OpenCV

项目需要做人脸识别，通过人脸进行自动切图，开源项目就OpenCV比较全的文档，且有Java的实现方案，所以就使用OpenCV进行开发。不过OpenCV进行人脸识别需要安装很多库文件和依赖，过程繁琐且复杂，为了避免再次入坑，下面就把这个过程记录下来，以备不时之需。

准备：以下需要提前做，否则后面会报错：

1、升级gcc版本:

# curl -Lks http://www.hop5.in/yum/el6/hop5.repo > /etc/yum.repos.d/hop5.repo

# yum install gcc gcc-g++ -y

# gcc --version

# g++ --version

1. 安装依赖

Java--即JDK，此处略

yum -y install epel-release

rpm --import http://li.nux.ro/download/nux/RPM-GPG-KEY-nux.ro

rpm -Uvh http://li.nux.ro/download/nux/dextop/el6/x86_64/nux-dextop-release-0-2.el6.nux.noarch.rpm

yum install -y gcc gcc-c++ make cmake ant ImageMagick ImageMagick-devel GraphicsMagick autoconf automake libpng-devel libjpeg-devel libtiff-devel jasper-devel freetype-devel libtool-ltdl libtool-ltdl-devel* libtool-ltdl libtool-ltdl-devel libgnomeui-devel

2.ant 验证

ant --help

3. ImageMagick验证

convert -version

4. GraphicsMagick验证

gm

或者

gm convert -list formats

安装OpenCV

1. 安装OpenCV依赖库（可能和前面yum安装有重复，忽略）

yum install -y gcc gcc-c++ cmake camke-gui git pkgconfig gtk2 gtk2-devel gtk2-devel-docs python python-pip python-devel python-numpy numpy gimp-devel gimp-devel-tools gimp-help-browser libjpeg-devel libtiff-devel libpng-devel gstreamer-devel gstreamer-plugins-base-devel libv4l libv4l-devel mencoder flvtool2 libavc1394-devel libraw1394-devel libdc1394-devel jasper-devel jasper-utils zlib-devel unzip swig libtool nasm libgnomeui-devel gnome-devel gnome-devel-docs kernel-devel gcc-gfortran ffmpeg ffmpeg-devel

升级 binutils至2.25

#wget http://ftp.gnu.org/gnu/binutils/binutils-2.25.1.tar.bz2

#tar -xjf binutils-2.25.1.tar.bz2

# cd binutils-2.25.1

# ./configure --prefix=/usr

# make

# make install

# as --version

# objdump -v

# ld -v

#若安装 ffmpeg-devel flvtool2 失败，记得执行以下步骤：

rpm --import http://li.nux.ro/download/nux/RPM-GPG-KEY-nux.ro

rpm -Uvh http://li.nux.ro/download/nux/dextop/el6/x86_64/nux-dextop-release-0-2.el6.nux.noarch.rpm

2. 下载

可以到官网下载 选择对应版本，我这里是3.4.1

https://codeload.github.com/opencv/opencv/zip/3.4.1

wget https://codeload.github.com/opencv/opencv/zip/3.4.1 --no-check_certificate

3. 解压

先解压，然后进入解压目录，创建 build 目录，进入build 目录，准备预编译

unzip opencv-3.4.1.zip

cd opencv-3.4.1

mkdir build && cd build/

4. 预编译

通过cmake命令预先编译一次，编译完成后查看输出结果, 在To be built里包含java这一项就表示预编译成功

cmake -DCMAKE_BUILD_TYPE=Release -DBUILD_SHARED_LIBS=OFF -DBUILD_EXAMPLES=OFF -DBUILD_TESTS=OFF -DBUILD_PERF_TESTS=OFF -DBUILD_TIFF=ON -DWITH_GPHOTO2=OFF -DCMAKE_INSTALL_PREFIX=/usr/local ..

预编译过程中，在下载ippicv_linux_xxxxxxxx.tgz这个文件会慢，(这里有一个解决方案，就是先终止预编译，然后手动下载后，放入opencv-3.1.0/3rdparty/ippicv/downloads/linux-808b791a6eac9ed78d32a7666804320e/ 目录, 然后去CSDN下载地址 下载， 也可以去官网获取下载地址，然后用迅雷下。 注意：linux-808b791a6eac9ed78d32a7666804320e目录名称是根据你的机器随机变化的，每个人的服务器上 linux- 后面可能不一样.)

5. 编译

cmake完成后，在build目录直接输入make进行编译，建议不使用-j n，多线程编译可能会出问题。

make

############################################

编译过程可能出现异常，提示：

1、

假如没有更新gcc版本，则会报如下错误:

In file included from /usr/local/src/opencv-3.4.1/3rdparty/include/opencl/1.2/CL/cl.h:30,

from /usr/local/src/opencv-3.4.1/modules/core/include/opencv2/core/opencl/runtime/autogenerated/opencl_core.hpp:101,

from /usr/local/src/opencv-3.4.1/modules/core/include/opencv2/core/opencl/runtime/opencl_core.hpp:68,

from /usr/local/src/opencv-3.4.1/modules/core/src/va_intel.cpp:22:

/usr/local/src/opencv-3.4.1/3rdparty/include/opencl/1.2/CL/cl_platform.h:214: error: reference to ‘int8_t’ is ambiguous

/usr/include/stdint.h:37: error: candidates are: typedef signed char int8_t

/usr/local/src/opencv-3.4.1/modules/core/include/opencv2/core/cvdef.h:486: error: typedef int8_t cv::int8_t

/usr/local/src/opencv-3.4.1/3rdparty/include/opencl/1.2/CL/cl_platform.h:214: error: reference to ‘int8_t’ is ambiguous

/usr/include/stdint.h:37: error: candidates are: typedef signed char int8_t

/usr/local/src/opencv-3.4.1/modules/core/include/opencv2/core/cvdef.h:486: error: typedef int8_t cv::int8_t

……………………

/usr/local/src/opencv-3.4.1/modules/core/include/opencv2/core/opencl/runtime/autogenerated/opencl_core.hpp:370: error: redefinition of ‘int cl_int’

/usr/local/src/opencv-3.4.1/modules/core/include/opencv2/core/opencl/runtime/autogenerated/opencl_core.hpp:283: error: ‘int cl_int’ previously defined here

/usr/local/src/opencv-3.4.1/modules/core/include/opencv2/core/opencl/runtime/autogenerated/opencl_core.hpp:370: error: ‘clWaitForEvents_pfn’ was not declared in this scope

make[2]: *** [modules/core/CMakeFiles/opencv_core.dir/src/va_intel.cpp.o] Error 1

make[1]: *** [modules/core/CMakeFiles/opencv_core.dir/all] Error 2

make: *** [all] Error 2

解决办法：

升级gcc版本至4.8，重新预编译：cmake后，再make。

2、

假如没有升级binutils的版本

[ 49%] Building CXX object modules/core/CMakeFiles/opencv_core.dir/src/convert.avx2.cpp.o

/tmp/ccA6i8a9.s: Assembler messages:

/tmp/ccA6i8a9.s:11: Error: suffix or operands invalid for `vbroadcastss'

/tmp/ccA6i8a9.s:12: Error: suffix or operands invalid for `vbroadcastss'

/tmp/ccA6i8a9.s:27: Error: no such instruction: `vpermq $216,%ymm2,%ymm2'

/tmp/ccA6i8a9.s:28: Error: suffix or operands invalid for `vpunpcklwd'

/tmp/ccA6i8a9.s:29: Error: suffix or operands invalid for `vpunpckhwd'

/tmp/ccA6i8a9.s:30: Error: suffix or operands invalid for `vpsrad'

/tmp/ccA6i8a9.s:33: Error: suffix or operands invalid for `vpsrad'

make[2]: *** [modules/core/CMakeFiles/opencv_core.dir/src/convert.avx2.cpp.o] Error 1

make[1]: *** [modules/core/CMakeFiles/opencv_core.dir/all] Error 2

make: *** [all] Error 2

解决办法：

升级 binutils至2.25

#wget http://ftp.gnu.org/gnu/binutils/binutils-2.25.1.tar.bz2

#tar -xjf binutils-2.25.1.tar.bz2

# cd binutils-2.25.1

# ./configure --prefix=/usr

# make

# make install

# as --version

# objdump -v

# ld -v

3、

/usr/include/jasper/jas_math.h:117:22: error: ‘SIZE_MAX’ was not declared in this scope

出现这个异常，不要慌，可以通过修改/usr/include/jasper/jas_math.h 文件源码解决。

在/usr/include/jasper/jas_math.h 的头部#include 的下面添加:

#if ! defined SIZE_MAX

#endif

然后重新编译就可以了。如果不行，请删除build目录下的内容，重新预编译、编译就应该没问题了

4、

假如没有升级libgphonto2，会报如下错误：

[ 80%] Building CXX object modules/videoio/CMakeFiles/opencv_videoio.dir/src/cap_gphoto2.cpp.o

/usr/local/src/opencv-3.4.1/modules/videoio/src/cap_gphoto2.cpp: In member function ‘void cv::gphoto2::DigitalCameraCapture::initContext()’:

/usr/local/src/opencv-3.4.1/modules/videoio/src/cap_gphoto2.cpp:323:66: error: invalid conversion from ‘void (*)(GPContext*, const char*, void*) ’ to ‘GPContextErrorFunc ’ [-fpermissive]

gp_context_set_error_func(context, ctxErrorFunc, (void*) this);

………………………………

/usr/local/src/opencv-3.4.1/modules/videoio/src/cap_gphoto2.cpp:248:27: note: in definition of macro ‘CR’

if ((r_0629c47b758 = (GPHOTO2_FUN))

^

make[2]: *** [modules/videoio/CMakeFiles/opencv_videoio.dir/src/cap_gphoto2.cpp.o] Error 1

make[1]: *** [modules/videoio/CMakeFiles/opencv_videoio.dir/all] Error 2

make: *** [all] Error 2

解决办法：编译升级 libgphonto2 phonto2 至2.5版本（未验证） 或 cmake时 加入-DWITH_GPHOTO2=OFF

wget https://nchc.dl.sourceforge.net/project/gphoto/libgphoto/2.5.8/libgphoto2-2.5.8.tar.bz2 --no-check_certificate

wget https://nchc.dl.sourceforge.net/project/gphoto/gphoto/2.5.8/gphoto2-2.5.8.tar.bz2 --no-check_certificate

5、没有升级python2.7 会报如下错误：

[ 91%] Generate files for Java bindings

File "/usr/local/src/opencv-3.4.1/modules/java/generator/../generator/gen_java.py", line 1079

^

SyntaxError: invalid syntax

make[2]: *** [CMakeFiles/dephelper/gen_opencv_java_source] Error 1

make[1]: *** [modules/java_bindings_generator/CMakeFiles/gen_opencv_java_source.dir/all] Error 2

make: *** [all] Error 2

解决办法：升级python至2.7，不要用3，

# wget https://www.python.org/ftp/python/2.7.11/Python-2.7.11.tar.xz ; --no-check-certificate

# xz -d Python-2.7.11.tar.xz

# tar -xf Python-2.7.11.tar

# cd Python-2.7.11

# ./configure

# make

# make install

# mv /usr/bin/python /usr/bin/python.bak

# ln -s /usr/local/bin/python2.7 /usr/bin/python

# vi /usr/bin/yum

!/usr/bin/python，修改成：!/usr/bin/python2.6

###############################################################################################################

6. 安装

编译完成，就可以进行安装，安装过程是生成opencv对应的库文件，我这里是java项目需要用到，所以也会顺便生成java相关的库文件

make install

命令执行完成，会在结果中看到/usr/local/share/OpenCV/java/目录生成了两个java相关的依赖库文件 opencv-341.jar 和 libopencv_java341.so ， 当然，每个人的环境不同，可能目录也不同，具体看日志输出就行了。

# ll /usr/local/share/OpenCV/java/ -h

total 69M

-rwxr-xr-x 1 root root 68M May 24 22:57 libopencv_java341.so

-rw-r--r-- 1 root root 413K May 24 22:54 opencv-341.jar

测试

先略，后期添加