前言:花了约三周看文档(打酱油),又花了两周搭环境,终于把tesseract用起来了,对简体中文的识别率还不错,在95%以上。现在简要记录一下安装、识别过程。
一、系统环境
系统:centos6.5
编译环境:g++
依赖软件:leptonica、opencv2.4.9、tesseract3.02
二、安装过程
(1) leptonica
sudo yum -y install autoconf automake libtool
sudo yum -y install autoconf-archive
sudo yum -y install pkgconfig
sudo yum -y install libpng12-dev
sudo yum -y install libjpeg8-dev
sudo yum -y install libtiff5-dev
sudo yum -y install zlib1g-dev
wget http://www.leptonica.org/source/leptonica-1.68.tar.gz tar xvzf leptonica-1.68.tar.gz cd leptonica-1.68/ ./configure make && make install
(2) tesseract3.02
tesseract的安装参考这里
同时参考官网这里
./autogen.sh
./configure --enable-debug LDFLAGS="-L/usr/local/lib" CFLAGS="-I/usr/local/include" make
make install
ldconfig
语言文件:
export TESSDATA_PREFIX=/some/path/to/tessdata
to point to your tessdata directory (example: if your tessdata path is ‘/usr/local/share/tessdata‘ you have to use ‘export TESSDATA_PREFIX=‘/usr/local/share/‘).
环境变量TESSDATA_PREFIX的路径需要设置成为tessdata文件夹的父目录。
(3) opencv2.4.9
- $ sudo yum -y install gtk2-devel tbb-devel libpng-devel
- $ wget http://sourceforge.net/projects/opencvlibrary/files/opencv-unix/2.4.9/opencv-2.4.9.zip
- $ unzip opencv-2.4.9.zip
- $ cd opencv-2.4.9
- $ mkdir build
- $ cd build
- $ cmake -D CMAKE_BUILD_TYPE=RELEASE -D CMAKE_INSTALL_PREFIX=/usr/local ..
- $ make -j2
- $ make install
三、API接口应用过程
(1) 编译过程
1、设置PKG_CONFIG_PATH environment variable ,加入`tesseract.pc‘
$echo ‘export PKG_CONFIG_PATH=$PKG_CONFIG_PATH:/usr/local/lib/pkgconfig‘ >> ~/.bashrc
$source ~/.bashrc
2、提示opencv缺libcufft,libnpps,libnppi,libnppc,libcudart等几个库,可以参考这里
这些库在cuda/lib64中,建立软连接
[[email protected] lib64]#ln -s /usr/local/cuda-6.5/lib64/libcufft.so.6.5 /usr/local/lib/libcufft.so
[[email protected] lib64]#ln -s /usr/local/cuda-6.5/lib64/libnpps.so.6.5 /usr/local/lib/libnpps.so
[[email protected] lib64]# ln -s /usr/local/cuda-6.5/lib64/libnppi.so.6.5 /usr/local/lib/libnppi.so
[[email protected] lib64]# ln -s /usr/local/cuda-6.5/lib64/libnppc.so.6.5 /usr/local/lib/libnppc.so
[[email protected] lib64]# ln -s /usr/local/cuda-6.5/lib64/libcudart.so.6.5 /usr/local/lib/libcudart.so
在运行时,报错:
error while loading shared libraries: libcufft.so.6.5: cannot open shared object file: No such file or directory
error while loading shared libraries: libnpps.so.6.5: cannot open shared object file: No such file or directory
error while loading shared libraries: libnppi.so.6.5: cannot open shared object file: No such file or directory
error while loading shared libraries: libnppc.so.6.5: cannot open shared object file: No such file or directory
error while loading shared libraries: libcudart.so.6.5: cannot open shared object file: No such file or directory
解决方法,参考这里
When I run testing routine, facing error: error while loading shared libraries: libcudart.so.6.5: cannot open shared object file: No such file or directory.
Solution for this, copy respect library to /usr/local/lib:
sudo cp /usr/local/cuda-6.5/lib64/libcudart.so.6.5 /usr/local/lib/libcudart.so.6.5 && sudo ldconfig
sudo cp /usr/local/cuda-6.5/lib64/libcublas.so.6.5 /usr/local/lib/libcublas.so.6.5 && sudo ldconfig
sudo cp /usr/local/cuda-6.5/lib64/libcurand.so.6.5 /usr/local/lib/libcurand.so.6.5 && sudo ldconfig
最后的结果是能够成功进行识别,但是准确率较windows下有所下降,唯一的区别是,linux用的opencv是2.4.9,而windows中使用的2.4.10.