Home
Softono

ImageOCR

Open source MIT PHP
269
Stars
67
Forks
2
Issues
14
Watchers
7 years
Last Commit

 About ImageOCR

ImageOCR is a PHP CAPTCHA recognition library designed to identify CAPTCHA images. It works well with non-overlapping characters and handles slightly overlapping characters with reasonable accuracy. The recognition process follows this pipeline: initialization, grayscale conversion, binarization, noise removal, segmentation, standardization, and recognition. Key features include three binarization modes: fixed threshold, dynamic background-based thresholding with maximum, minimum, and background options. Noise removal supports two methods: isolated spot removal and connected domain removal. Segmentation supports equal-width splitting, connected domain splitting for non-overlapping strings, and a drip algorithm for overlapping characters. Standardization normalizes images to configurable width and height. Debug mode is available to visualize intermediate results. A bundled Docker image provides a web interface for quick experimentation. The library can be installed via Composer (mohuishou/image-ocr) and expose

Platforms

Web Self-hosted

Languages

PHP

Links

Need Help Installing ImageOCR?

We provide expert installation service for this software. Our team will install, configure, and secure ImageOCR on your server. plans start at just $30.

ImageOCR

php 验证码识别库,对于非粘连字符具有很好的识别效果,对于一般粘连字符也能有较为良好的识别 除噪算法支持孤立点除杂和连通域除噪,分割算法支持等宽分割、连通域分割以及滴水算法分割

示例效果

示例

Install

composer require mohuishou/image-ocr

使用方法

例子详见 example

use docker

docker run --rm -p 8088:8088 mohuishou/image-ocr

点击 http://localhost:8088 查看效果

大致流程:

初始化 -> 灰度化 ---> 二值化 ---> 除噪点 -> 分割 -> 标准化 -> 识别

初始化

对象初始化

$image=new Image($img_path);
$image_ocr=new ImageOCR($image)

初始化二值化阈值

$image_ocr->setMaxGrey(90);
$image_ocr->setMinGrey(10);

初始化标准化图片宽高

$image_ocr->setStandardWidth(13);
$image_ocr->setStandardHeight(20);

开启 Debug

$image_ocr->setDebug(true);

灰度化

try{
    $image_ocr->grey();
}catch (\Exception $e){
    echo $e->getMessage();
}

二值化

注意:这一步的前提是需要先执行上一步灰度化,不然会抛出一个错误

try{
    $image_ocr->hash($max_grey=null,$min_grey=null);
}catch (\Exception $e){
    echo $e->getMessage();
}

二值化支持两种方式,第一种$image_ocr->hash($max_grey=null,$min_grey=null)即为上面那种固定的阈值范围,第二种为hashByBackground($model=self::MAX_MODEL,$max_grey=null,$min_grey=null),通过背景图像的灰度值,动态取阈值,支持三种模式MAX_MODEL,MIN_MODEL,BG_MODEL分别是最大值、最小值和背景模式,最大值模式会用背景的灰度值替换阈值的上限,最小值模式替换下限,背景模式上下限都替换,即为只去除背景

除噪点

前置条件为二值化

孤立点除噪法
try{
    $image_ocr->removeSpots();
}catch (\Exception $e){
    echo $e->getMessage();
}
连通域除噪法

[如果要使用连通域分割法,可以跳过连通域除噪点,分割的同时可以一并除噪]

try{
    //使用之前需要初始化连通域对象
    $image_ocr->setImageConnect();
    //除噪
    $image_ocr->removeSpotsByConnect();
}catch (\Exception $e){
    echo $e->getMessage();
}

分割

非粘连字符串

连通域分割法

try{
    //使用之前需要初始化连通域对象
    $image_ocr->setImageConnect();
    //分割
    $image_ocr->splitByConnect();
}catch (\Exception $e){
    echo $e->getMessage();
}
粘连字符串

滴水算法分割

TODO: 待测试

标准化

try{
    $standard_data=$image_ocr->standard();
}catch (\Exception $e){
    echo $e->getMessage();
}

识别

TODO:待完善

API

ImageOCR::__construct(Image $image)
ImageOCR::saveImage($path)
ImageOCR::grey()
ImageOCR::hash($max_grey=null,$min_grey=null)
ImageOCR::hashByBackground($model=self::MAX_MODEL,$max_grey=null,$min_grey=null)
ImageOCR::removeSpots()
ImageOCR::removeSpotsByConnect()
ImageOCR::standard()
ImageOCR::setImageConnect()
ImageOCR::setImage(Image $image)
ImageOCR::getStandardData()
ImageOCR::setMaxGrey($max_grey)
ImageOCR::setMinGrey($min_grey)
ImageOCR::setStandardWidth($standard_width)
ImageOCR::setStandardHeight($standard_height)

//ImageTool的方法均为静态方法
ImageTool::removeZero($data)
ImageTool::removeZeroColumn($hash_data)
ImageTool::drawBrowser($data)
ImageTool::transposeAndRemoveZero($hash_data)
ImageTool::hashTranspose($hash_data)
ImageTool::img2hash($img)
ImageTool::hash2img($hash_data,$padding=0)

CHANGELOG

0.2 [2017-4-1]

0.1 [2016-10-7]

  1. 默认模板保存方式由数据库改为文件,保存路径为./db/db.json
  2. 使用 composer 安装