一. 编译tesseract的系统环境：
mac系统版本：OS X Yosemite 10.10
Xcode版本：Version 6.2 (6C131e)
iOS SDK：8.0
二. 准备工作：
1、下载源代码
1.tesseract (https://code.google.com/p/tesseract-ocr/downloads/list)
2.tesseract语言包
3.leptonica(tesseract依赖于它) (http://www.leptonica.com/download.html)
2、新建文件夹 compile_code，讲上面下载的tesseract，leptonica都解压在这里
三. 开始编译：
开发环境配置有很多种，我使用的是MacPort的方式（主要是我第一个接触的就是它，懒得去换了）
1.下载MacPort（http://www.macports.org/install.php）
2.修改/opt/local/etc/macports/sources.conf文件
官网修改方案：http://trac.macports.org/wiki/howto/PortTreeTarball
sudo vim /opt/local/etc/macports/sources.conf找到：rsync://rsync.macports.org/release/ports/[default]
修改为：https://distfiles.macports.org/ports.tar.gz [default]
打开终端，输入下面的命令，等待更新结束
sudo port –d sync
3.按顺序安装所需要的工具：M4-> autoconf -> automake -> libtool
以M4安装为例：
sudo port install M4
4.进入文件夹compile_code，新建脚本文件build.sh
内容如下：
#!/bin/sh
GLOBAL_OUTDIR="`pwd`/build"
LOCAL_OUTDIR="./outdir"
LEPTON_LIB="`pwd`/leptonica-1.72"
TESSERACT_LIB="`pwd`/tesseract-ocr"
IOS_BASE_SDK="8.0"
IOS_DEPLOY_TGT="8.0"
# 设置编译器、连接器
export CXX=`xcrun -find c++`
export CC=`xcrun -find cc`
export LD=`xcrun -find ld`
export AR=`xcrun -find ar`
export AS=`xcrun -find as`
export NM=`xcrun -find nm`
export RANLIB=`xcrun -find ranlib`
# 设置编译器路径、模拟器路径
XCODE_DEVELOPER_PATH=/Applications/Xcode.app/Contents/Developer
XCODETOOLCHAIN_PATH=$XCODE_DEVELOPER_PATH/Toolchains/XcodeDefault.xctoolchain
SDK_IPHONEOS_PATH=$(xcrun --sdk iphoneos --show-sdk-path)
SDK_IPHONESIMULATOR_PATH=$(xcrun --sdk iphonesimulator --show-sdk-path)
export PATH="$XCODETOOLCHAIN_PATH/usr/bin:$PATH"
declare -a archs
archs=(arm7 arm7s arm64 i386 x86_64)
declare -a arch_name
arch_names=(arm-apple-darwin7 arm-apple-darwin7s arm-apple-darwin64 i386-apple-darwin x86_64-apple-darwin)
# 设置编译参数
setenv_all() {
# Add internal libs
export CFLAGS="$CFLAGS -I$GLOBAL_OUTDIR/include -L$GLOBAL_OUTDIR/lib -Qunused-arguments"
export LDFLAGS="-L$SDKROOT/usr/lib/"
export CPPFLAGS=$CFLAGS
export CXXFLAGS=$CFLAGS
}
setenv_arm7() {
unset DEVROOT SDKROOT CFLAGS CPP CXXCPP LDFLAGS CPPFLAGS CXXFLAGS
export SDKROOT=$SDK_IPHONEOS_PATH
export CFLAGS="-arch armv7 -pipe -no-cpp-precomp -isysroot $SDKROOT -miphoneos-version-min=$IOS_DEPLOY_TGT -I$SDKROOT/usr/include/"
setenv_all
}
setenv_arm7s() {
unset DEVROOT SDKROOT CFLAGS CPP CXXCPP LDFLAGS CPPFLAGS CXXFLAGS
export SDKROOT=$SDK_IPHONEOS_PATH
export CFLAGS="-arch armv7s -pipe -no-cpp-precomp -isysroot $SDKROOT -miphoneos-version-min=$IOS_DEPLOY_TGT -I$SDKROOT/usr/include/"
setenv_all
}
setenv_arm64() {
unset DEVROOT SDKROOT CFLAGS CPP CXXCPP LDFLAGS CPPFLAGS CXXFLAGS
export SDKROOT=$SDK_IPHONEOS_PATH
export CFLAGS="-arch arm64 -pipe -no-cpp-precomp -isysroot $SDKROOT -miphoneos-version-min=$IOS_DEPLOY_TGT -I$SDKROOT/usr/include/"
setenv_all
}
setenv_i386() {
unset DEVROOT SDKROOT CFLAGS CPP CXXCPP LDFLAGS CPPFLAGS CXXFLAGS
export SDKROOT=$SDK_IPHONESIMULATOR_PATH
export CFLAGS="-arch i386 -pipe -no-cpp-precomp -isysroot $SDKROOT -miphoneos-version-min=$IOS_DEPLOY_TGT"
setenv_all
}
setenv_x86_64() {
unset DEVROOT SDKROOT CFLAGS CPP CXXCPP LDFLAGS CPPFLAGS CXXFLAGS
export SDKROOT=$SDK_IPHONESIMULATOR_PATH
export CFLAGS="-arch x86_64 -pipe -no-cpp-precomp -isysroot $SDKROOT -miphoneos-version-min=$IOS_DEPLOY_TGT"
setenv_all
}
# 合并库（用于模拟器与真机同时使用）
create_outdir_lipo() {
for file in `find $LOCAL_OUTDIR/i386 -name "lib*.a"`; do
lib_arm7=`echo $file | sed "s/i386/arm7/g"`
lib_arm7s=`echo $file | sed "s/i386/arm7s/g"`
lib_arm64=`echo $file | sed "s/i386/arm64/g"`
lib_x86_64=`echo $file | sed "s/i386/x86_64/g"`
lib_i386=`echo $file`
lib=`echo $file | sed "s/i386//g"`
xcrun -sdk iphoneos lipo -arch armv7s $lib_arm7s -arch armv7 $lib_arm7 -arch arm64 $lib_arm64 -arch i386 $lib_i386 -arch x86_64 $lib_x86_64 -create -output $lib
done
}
merge_libfiles() {
DIR=$1
LIBNAME=$2
cd $DIR
for i in `find . -name "lib*.a"`; do
$AR -x $i
done
$AR -r $LIBNAME *.o
rm -rf *.o __*
cd -
}
# 下载源代码
#get_leptonica()
#{
# curl -C - -O http://www.leptonica.com/source/leptonica-1.68.tar.gz
# tar xzvf leptonica-1.68.tar.gz
#}
#get_tessreact()
#{
# curl -C - -O http://tesseract-ocr.googlecode.com/files/tesseract-3.01.tar.gz
# tar xzvf tesseract-3.01.tar.gz
#}
#######################
# 清理工作空间
#######################
rm -rf $GLOBAL_OUTDIR lib include
########################
## 下载需要编译的文件
########################
## 下载leptonica
#get_leptonica
## 下载tessreact
#get_tessreact
#######################
# 进入LEPTONLIB目录
#######################
cd $LEPTON_LIB
rm -rf $LOCAL_OUTDIR
for n in "${!archs[@]}"
do
mkdir -p "$LOCAL_OUTDIR/${archs[$n]}"
make clean 2> /dev/null
make distclean 2> /dev/null
eval "setenv_${archs[$n]}"
./configure --host="${arch_names[$n]}" --enable-shared=no --disable-programs --without-zlib --without-libpng --without-jpeg --without-giflib --without-libtiff
make -j12
cp -rvf src/.libs/lib*.a "$LOCAL_OUTDIR/${archs[$n]}"
done
create_outdir_lipo
mkdir -p $GLOBAL_OUTDIR/include/leptonica && find ./ -name '*.h' -exec cp {} $GLOBAL_OUTDIR/include/leptonica/ \;
mkdir -p $GLOBAL_OUTDIR/lib && cp -rvf $LOCAL_OUTDIR/lib*.a $GLOBAL_OUTDIR/lib
cd ..
#######################
# 进入TESSERACT目录
#######################
cd $TESSERACT_LIB
rm -rf $LOCAL_OUTDIR
for n in "${!archs[@]}"
do
mkdir -p "$LOCAL_OUTDIR/${archs[$n]}"
make clean 2> /dev/null
make distclean 2> /dev/null
eval "setenv_${archs[$n]}"
bash autogen.sh
LIBLEPT_HEADERSDIR=$GLOBAL_OUTDIR/include ./configure --host="${arch_names[$n]}" --enable-shared=no --disable-graphics
make -j12
for i in `find . -name "lib*.a" | grep -v $LOCAL_OUTDIR`
do
cp -rvf $i "$LOCAL_OUTDIR/${archs[$n]}"
done
merge_libfiles "$LOCAL_OUTDIR/${archs[$n]}" libtesseract_all.a
done
create_outdir_lipo
mkdir -p $GLOBAL_OUTDIR/include/tesseract && find ./ -name '*.h' -exec cp {} $GLOBAL_OUTDIR/include/tesseract/ \;
mkdir -p $GLOBAL_OUTDIR/lib && cp -rvf $LOCAL_OUTDIR/lib*.a $GLOBAL_OUTDIR/lib
make clean 2> /dev/null
make distclean 2> /dev/null
cd ..
#######################
# Copying
#######################
cp -rf $GLOBAL_OUTDIR/include .
mkdir -p lib
cp -rf $GLOBAL_OUTDIR/lib/libtesseract_all.a $GLOBAL_OUTDIR/lib/liblept.a lib/
echo "Finished!"
6.为build.sh增加可执行权限：
chmod a+x build.sh
7.终端中运行build.sh
运行结束后会在文件夹compile_code中的lib文件夹中生成两个*.a文件，include文件夹中生成头文件，这两个文件夹引入工程中即可使用该框架了。
四. iOS上的tesseract实现
/*
* tesseract.h(来源网上)
*/
#import <Foundation/Foundation.h>
#import <UIKit/UIKit.h>
@interface Tesseract : NSObject {
// OCR识别库
NSString* _dataPath;
// 需要识别的语言
NSString* _language;
// tesseract参数
NSMutableDictionary* _variables;
}
+ (NSString *)version;
/**
* 设置OCR识别参数
*
* @param dataPath OCR识别库
* @param language 所选用的语言
*/
- (id)initWithDataPath:(NSString *)dataPath language:(NSString *)language;
- (void)setVariableValue:(NSString *)value forKey:(NSString *)key;
/**
* 设置识别的图片
*
* @param image 需要识别的图片
*/
- (void)setImage:(UIImage *)image;
/**
* OCR识别语言设置
*
* @param language 需要识别的语言
*/
- (BOOL)setLanguage:(NSString *)language;
- (BOOL)recognize;
- (NSString *)recognizedText;
@end/*
* tesseract.mm
*/
#import "Tesseract.h"
#import "baseapi.h"
#import "environ.h"
#import "pix.h"
namespace tesseract {
class TessBaseAPI;
};
@interface Tesseract () {
tesseract::TessBaseAPI* _tesseract;
// 转换为二进制的图片数据
uint32_t* _pixels;
}
@end
@implementation Tesseract
/**
* 获取版本信息
*/
+ (NSString *)version {
return [NSString stringWithFormat:@"%s", tesseract::TessBaseAPI::Version()];
}
/**
* 设置参数
*
* @param dataPath 黑白名单设置
* @param language 所选用的语言
*/
- (id)initWithDataPath:(NSString *)dataPath language:(NSString *)language {
self = [super init];
if (self) {
_dataPath = dataPath;
_language = language;
_variables = [[NSMutableDictionary alloc] init];
[self copyDataToDocumentsDirectory];
_tesseract = new tesseract::TessBaseAPI();
BOOL success = [self initEngine];
if (!success) {
return nil;
}
}
return self;
}
/**
* 初始化tesseract识别
*/
- (BOOL)initEngine {
int returnCode = _tesseract->Init([_dataPath UTF8String], [_language UTF8String]);
return (returnCode == 0) ? YES : NO;
}
/**
* 将识别库拷贝至 Documents 文件夹
*/
- (void)copyDataToDocumentsDirectory {
// Useful paths
NSFileManager *fileManager = [NSFileManager defaultManager];
NSArray *documentPaths = NSSearchPathForDirectoriesInDomains(NSDocumentDirectory, NSUserDomainMask, YES);
NSString *documentPath = ([documentPaths count] > 0) ? [documentPaths objectAtIndex:0] : nil;
NSString *dataPath = [documentPath stringByAppendingPathComponent:_dataPath];
// Copy data in Doc Directory
if (![fileManager fileExistsAtPath:dataPath]) {
NSString *bundlePath = [[NSBundle bundleForClass:[self class]] bundlePath];
NSString *tessdataPath = [bundlePath stringByAppendingPathComponent:_dataPath];
if (tessdataPath) {
[fileManager createDirectoryAtPath:documentPath withIntermediateDirectories:YES attributes:nil error:NULL];
[fileManager copyItemAtPath:tessdataPath toPath:dataPath error:nil];
}
}
setenv("TESSDATA_PREFIX", [[documentPath stringByAppendingString:@"/"] UTF8String], 1);
}
/**
* 设置OCR识别参数
*
* @param value OCR识别参数属性
* @param key OCR识别参数键值(参考Tesseract源代码：contrib/dump.config)
*/
- (void)setVariableValue:(NSString *)value forKey:(NSString *)key {
/*
* Example:
* _tesseract->SetVariable("tessedit_char_whitelist", "0123456789ABCDEFGHIJKLMNOPQRSTUVWXYZ");
* _tesseract->SetVariable("language_model_penalty_non_freq_dict_word", "0");
* _tesseract->SetVariable("language_model_penalty_non_dict_word ", "0");
*/
[_variables setValue:value forKey:key];
_tesseract->SetVariable([key UTF8String], [value UTF8String]);
}
/**
* 重新设置OCR识别参数
*/
- (void)loadVariables {
for (NSString* key in _variables) {
NSString* value = [_variables objectForKey:key];
_tesseract->SetVariable([key UTF8String], [value UTF8String]);
}
}
/**
* 设置识别语言
*/
- (BOOL)setLanguage:(NSString *)language {
_language = language;
// int returnCode = [self initEngine];
// if (returnCode != 0) return NO;
/*
* "WARNING: On changing languages, all Tesseract parameters
* are reset back to their default values."
*/
[self loadVariables];
return YES;
}
/**
* OCR识别
*
* @return 识别是否成功
*/
- (BOOL)recognize {
int returnCode = _tesseract->Recognize(NULL);
return (returnCode == 0) ? YES : NO;
}
/**
* OCR识别
*
* @return 识别出来的文本
*/
- (NSString *)recognizedText {
char* utf8Text = _tesseract->GetUTF8Text();
return [NSString stringWithUTF8String:utf8Text];
}
/**
* 设置识别的图片
*
* @param image 需要识别的图片
*/
- (void)setImage:(UIImage *)image
{
free(_pixels);
CGSize size = [image size];
int width = size.width;
int height = size.height;
if (width <= 0 || height <= 0) {
return;
}
_pixels = (uint32_t *) malloc(width * height * sizeof(uint32_t));
// 清空图片缓存数组
memset(_pixels, 0, width * height * sizeof(uint32_t));
// 申请RGA图片缓存
CGColorSpaceRef colorSpace = CGColorSpaceCreateDeviceRGB();
// 申请 _pixels 图片上下文
CGContextRef context = CGBitmapContextCreate(_pixels, width, height, 8, width * sizeof(uint32_t), colorSpace, kCGBitmapByteOrder32Little | kCGImageAlphaPremultipliedLast);
// 向上下文填充二进制图片数据(该数据会填充到_pixels数组中)
CGContextDrawImage(context, CGRectMake(0, 0, width, height), [image CGImage]);
// 释放上下文与颜色空间
CGContextRelease(context);
CGColorSpaceRelease(colorSpace);
// 传递图片给 Tesseract 库
_tesseract->SetImage((const unsigned char *) _pixels, width, height, sizeof(uint32_t), width * sizeof(uint32_t));
}
-(void)dealloc{
// 关闭tesseract
_tesseract->End();
}
@end
Tesseract* tesseract = [[Tesseract alloc] initWithDataPath:@"tessdata" language:@"eng"];
self.tesseract = tesseract;
// tesseract参数
// [tesseract setVariableValue:@"textord_words_default_minspace" forKey:@"0.2"];
// [tesseract setVariableValue:@"textord_words_width_ild" forKey:@"0.4"];
// 设置需要识别的图片
UIImage *image = [UIImage imageNamed:@"Capture2.png"];
UIImage *conImage = image;
// 图像预处理
conImage = [UIImage convertToGrayscale:conImage];
conImage = [conImage sharpen];
conImage = [UIImage grayImage:conImage];
// 显示图片
self.imageView.image = image;
self.conImageView.image = conImage;
[self.tesseract setImage:conImage];
[self.tesseract recognize];
NSLog(@"%@", [self.tesseract recognizedText]);
[self.tesseract setImage:nil];
五. tesseract小结
网上的一些资料都感觉很古老，很陈旧了，编写这部分代码什么的也弄了2天多的时间。程序现在感觉还是有些问题，但是起码可以识别一些效果比较好的英文了。有时间还会继续研究。个人认为，后续的研究应该就是图像预处理的部分了，有熟悉的大师也可以帮助一下。互勉，共同进步。
网上感觉不错的例子：https://github.com/gali8/Tesseract-OCR-iOS或者http://code4app.com/ios/Tesseract-OCR/533d7ada933bf024048b4c8b
一些简单的图像处理函数：http://pan.baidu.com/share/link?shareid=1816900461&uk=218547027（自己网上摘的）
按照网上的说法，tesseract识别的关键点在于字典训练还有图片预处理，因此后续研究方向估计就是图片预处理这部分了。
扫一扫
866
被折叠的 条评论
为什么被折叠?
到【灌水乐园】发言
