光学字符识别

Happy Monkey

于 2024-08-05 06:22:21 发布

阅读量405

点赞数 19

分类专栏：字体识别 OCR 自动识别文章标签：计算机视觉自动化人工智能机器学习图像处理

本文链接：https://blog.csdn.net/zhangyy57/article/details/140916474

版权

自动识别同时被 3 个专栏收录

32 篇文章 0 订阅

订阅专栏

字体识别

15 篇文章 0 订阅

订阅专栏

OCR

6 篇文章 0 订阅

订阅专栏

这段代码是一个使用HALCON软件进行OCR（光学字符识别）的示例脚本。
dev_update_off ()
dev_close_window ()

Initialize window handle
dev_open_window_fit_size (0, 0, 960, 768, 700, 700, WindowHandle)
set_display_font (WindowHandle, 16, ‘mono’, ‘true’, ‘false’)
Read OCR classifier and create the text model
read_ocr_class_mlp (‘Industrial_Rej’, OCRHandle)
Initialize display variables

disp_message (WindowHandle, ‘This example shows how find_text can be used to \nfind and read text. Furthermore, the most \nimportant parameters are explained.’, ‘window’, 12, 12, ‘black’, ‘true’)
disp_continue_message (WindowHandle, ‘black’, ‘true’)
stop ()
*
dev_clear_window ()
disp_message (WindowHandle, ‘Essentially, the only thing that needs to be \ndone before calling find_text is to create \na text model with create_text_model_reader \nwith mode set to ‘auto’.\n \nThe resulting segmentation can then be obtained \nby calling the operators get_text_result and \nget_text_object.’, ‘window’, 12, 12, ‘black’, ‘true’)
disp_continue_message (WindowHandle, ‘black’, ‘true’)
stop ()

An OCR Classifier based on a multilayer perceptron (MLP) is required
create_text_model_reader (‘auto’, OCRHandle, TextModel)
clear_ocr_class_mlp (OCRHandle)

read_image (Image, ‘ocr/medication_package_02_right’)
*

Without setting further parameters, find_text will find and read all
text within the input image
find_text (Image, TextModel, TextResult)
The segmented regions can be obtained by calling get_text_object
get_text_object (TextLines, TextResult, ‘all_lines’)

dev_set_colored (12)
dev_display (Image)
dev_display (TextLines)
disp_message (WindowHandle, ‘Find_text extracts all text within the input image,\nregardless of the character size. \nPer default, light and dark text is extracted.’, ‘window’, 12, 12, ‘black’, ‘true’)
disp_continue_message (WindowHandle, ‘black’, ‘true’)
clear_text_result (TextResult)
stop ()
*

It is possible to restrict the segmentation to text of a certain polarity
using the set_text_model_param operator
set_text_model_param (TextModel, ‘polarity’, ‘dark_on_light’)

find_text (Image, TextModel, TextResult)
*
get_text_object (TextLines, TextResult, ‘all_lines’)
dev_set_colored (12)
dev_display (Image)
dev_display (TextLines)
disp_message (WindowHandle, ‘It is possible to restrict the text segmentation \naccording to the polarity of the text. \n \nFor example, restricting the polarity to dark text \non a light background ignores all light text.’, ‘window’, 12, 12, ‘black’, ‘true’)
disp_continue_message (WindowHandle, ‘black’, ‘true’)
stop ()
clear_text_result (TextResult)
*

Furthermore, it is possible to restrict the segmentation to text attributes
such as the character height, width or stroke width
set_text_model_param (TextModel, ‘polarity’, ‘dark_on_light’)
set_text_model_param (TextModel, ‘min_char_height’, 20)

find_text (Image, TextModel, TextResult)
*
get_text_object (TextLines, TextResult, ‘all_lines’)
dev_set_colored (12)
dev_display (Image)
dev_display (TextLines)
disp_message (WindowHandle, ‘Furthermore, it is possible to restrict the segmentation \nto text attributes such as the character height, \nwidth or stroke width.\n \nFor example, setting the minimal characters height \nto 20px ignores all characters with a smaller height.’, ‘window’, 12, 12, ‘black’, ‘true’)
disp_continue_message (WindowHandle, ‘black’, ‘true’)
dev_display (TextLines)
stop ()
clear_text_result (TextResult)
*

When searching for specific text structures, it is possible to
set the corresponding structure with set_text_model_param
set_text_model_param (TextModel, ‘polarity’, ‘dark_on_light’)
set_text_model_param (TextModel, ‘min_char_height’, 20)
The separators which are added to the ‘text_line_separators’
need to be valid characters within the used classifier. Otherwise
they are ignored.
set_text_model_param (TextModel, ‘text_line_separators’, ‘/’)
set_text_model_param (TextModel, ‘text_line_structure’, ‘2 4’)

find_text (Image, TextModel, TextResult)
*
get_text_object (TextLines, TextResult, ‘all_lines’)
dev_set_colored (12)
dev_display (Image)
dev_display (TextLines)
disp_message (WindowHandle, ‘When searching for specific text structures, it can \nbe helpful to set the corresponding text_line_structure \nwith set_text_model_param. \n \nFor example, when searching for a date of the form MM/YYYY,\nit is possible to add ‘/’ to the text_line_separators and \nset the text_line_structure to ‘2 4’.’, ‘window’, 12, 12, ‘black’, ‘true’)
disp_continue_message (WindowHandle, ‘black’, ‘true’)
stop ()
*

It is possible to directly retain the classification results for
each of the segmented characters with the get_text_result operator

dev_set_colored (12)
dev_display (Image)
dev_display (TextLines)
*

Display the single characters
smallest_rectangle1 (TextLines, Row1, Column1, Row2, Column2)
get_text_result (TextResult, ‘class’, SingleCharacters)
tuple_sum (SingleCharacters, TextLineCharacters)
dev_set_color (‘dark green’)
for CharacterIndex := 0 to |SingleCharacters| - 1 by 1
set_tposition (WindowHandle, Row2[CharacterIndex] + 10, Column1[CharacterIndex])
write_string (WindowHandle, SingleCharacters[CharacterIndex])
endfor
disp_message (WindowHandle, 'It is possible to directly retain the classification \nresults for each of the segmented characters \nwith the get_text_result operator. \n \nExtracted Text: ’ + TextLineCharacters, ‘window’, 12, 12, ‘black’, ‘true’)
stop ()
clear_text_result (TextResult)
clear_text_model (TextModel)

medication_package_02_right.png
在这里插入图片描述
以下是程序运行结果：