图片识别之pytesser3使用

待我称王封你为后i 2022-02-20 21:13 389阅读 0赞

**一、安装：**

1、安装pytesser3（pip install pytesser3）；

2、下载安装Tesseract-OCR引擎，否则易出错。

安装过程按照下面的图片指示进行。注意选择安装语言包，包含英文（默认安装）、中文、数学公式等，可根据需要自行下载。

![watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3FxXzM4ODgyMzI3_size_16_color_FFFFFF_t_70][]

![watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3FxXzM4ODgyMzI3_size_16_color_FFFFFF_t_70 1][]

![watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3FxXzM4ODgyMzI3_size_16_color_FFFFFF_t_70 2][]

![watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3FxXzM4ODgyMzI3_size_16_color_FFFFFF_t_70 3][]

![watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3FxXzM4ODgyMzI3_size_16_color_FFFFFF_t_70 4][]

![watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3FxXzM4ODgyMzI3_size_16_color_FFFFFF_t_70 5][]

![watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3FxXzM4ODgyMzI3_size_16_color_FFFFFF_t_70 6][]

安装目录建议默认，以配合pytesser3里的\_\_init\_\_.py的下图路径

![20190410090101619.png][]

![watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3FxXzM4ODgyMzI3_size_16_color_FFFFFF_t_70 7][]

![watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3FxXzM4ODgyMzI3_size_16_color_FFFFFF_t_70 8][]

安装完成。

3、下载语言数据包[https://github.com/tesseract-ocr/tesseract/wiki/Data-Files][https_github.com_tesseract-ocr_tesseract_wiki_Data-Files]，并将数据包直接放置在C:\\Program Files (x86)\\Tesseract-OCR\\tessdata文件夹下（否则无法识别其他语言）

chi\_sim.traineddata中文简体

chi\_tra.traineddata中文繁体

4、识别其他语言：将pytesser3文件夹中\_\_init\_\_.py文件添加下图红框中的language参数，如下图：

![watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3FxXzM4ODgyMzI3_size_16_color_FFFFFF_t_70 9][]

将pytesser3文件夹中的util.py文件添加encoding=’utf-8’，否则可能会报错，如下图：

![watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3FxXzM4ODgyMzI3_size_16_color_FFFFFF_t_70 10][]

在调用image\_to\_string函数时，加上相应的language参数就可以了，简体中文为 chi\_sim， 繁体中文为chi\_tra，如image\_to\_string(img4, language='chi\_sim')。

**二、图片处理流程：**

1.  灰度处理：
    
        from PIL import Image
        
        img = Image.open(url)
        
        img.convert("L") #把彩色图像转化为灰度图像
        
        img.show()
    
     
2.  增强对比度：

from PIL import ImageEnhance
    
    img = ImageEnhance.Contrast(img)
    
    img = img.enhance(n)
    
    img.show()

**    3.二值化：**

pixdata = img.load()                         #加载图片数据
    
    w = img.size[0]                              #获取图片的宽度
    
    h = img.size[1]                              #获取图片的高度
    
    for x in range(w):
    
         for y in range(h):
    
              if pixdata[x, y] < threshold:
    
                   pixdata[x, y] = 0              #0为黑色
    
              else:
    
                   pixdata[x, y] = 255           #255为白色
    
    img.show()

4.降噪：

pixdata = img.load()
    
    w, h = img.size                             #获取图片宽高
    
    for x in range(1, w - 1):
    
        for y in range(1, h - 1):
    
            count = 0                          #初始化该像素点四周的白点个数
    
            if pixdata[x, y - 1] == 255:          #上
    
                count += 1
    
            if pixdata[x, y + 1] == 255:          #下
    
                count += 1
    
            if pixdata[x - 1, y] == 255:          #左
    
                count += 1
    
            if pixdata[x + 1, y] == 255:          #右，到此处截止为4邻域
    
                count += 1
    
            #加上下面4个方位共为8邻域，去噪点用4邻域即可
    
            if pixdata[x - 1, y - 1] == 255:       #左上
    
                count += 1
    
            if pixdata[x - 1, y + 1] == 255:       #左下
    
                count += 1
    
            if pixdata[x + 1, y - 1] == 255:       #右上
    
                count += 1
    
            if pixdata[x + 1, y + 1] == 255:      #右下，到此处截止为8邻域
    
                count += 1
    
            if count > 6:                      #若该像素点四周有4个白点，则
    
                pixdata[x, y] = 255             #将该像素点设为白色

**      5.图片切割：以后再补充。**

**      6.图片识别：**

#通过打开的文件识别
    
    image = Image.open('7.jpg')
    
    #图片识别，使用中文语言数据包时可能导致英文识别准确率降低
    
    print(pytesser3.image_to_string(image, language='chi_sim'))
    
    #通过文件路径直接识别
    
    print(pytesser3.image_file_to_string('7.jpg'))

[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3FxXzM4ODgyMzI3_size_16_color_FFFFFF_t_70]: /images/20220221/a4a5c02dc0424e509536ab6f30fd7084.png
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3FxXzM4ODgyMzI3_size_16_color_FFFFFF_t_70 1]: /images/20220221/4a0e5ad54dc94d26b2383c4153d32a74.png
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3FxXzM4ODgyMzI3_size_16_color_FFFFFF_t_70 2]: /images/20220221/32c1f43b016e42a79e2c2d3c53592077.png
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3FxXzM4ODgyMzI3_size_16_color_FFFFFF_t_70 3]: /images/20220221/b63016720a064078ab8b5f88039810fd.png
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3FxXzM4ODgyMzI3_size_16_color_FFFFFF_t_70 4]: /images/20220221/74d33fe852bb40ddb2b8610f8d67770d.png
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3FxXzM4ODgyMzI3_size_16_color_FFFFFF_t_70 5]: /images/20220221/027b81ac4fb34ceba751bd1e4c0086b4.png
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3FxXzM4ODgyMzI3_size_16_color_FFFFFF_t_70 6]: /images/20220221/7d5da111ffdc4ee7bbc40459deaf4208.png
[20190410090101619.png]: /images/20220221/6ccd472b6a1b433998622fab0961f9e9.png
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3FxXzM4ODgyMzI3_size_16_color_FFFFFF_t_70 7]: /images/20220221/4a7ee84adad74a9d8d31b915652fc761.png
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3FxXzM4ODgyMzI3_size_16_color_FFFFFF_t_70 8]: /images/20220221/e9c0e59b736946f3a0a303dff098eafb.png
[https_github.com_tesseract-ocr_tesseract_wiki_Data-Files]: https://github.com/tesseract-ocr/tesseract/wiki/Data-Files
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3FxXzM4ODgyMzI3_size_16_color_FFFFFF_t_70 9]: /images/20220221/d2309c8c99774440946d0a94126b340a.png
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3FxXzM4ODgyMzI3_size_16_color_FFFFFF_t_70 10]: /images/20220221/d9c3e7332bb34baf9e64aaa809130d6c.png