halcon基于mlp神经网络分类器的OCR字符识别 |
您所在的位置：网站首页 › 神经网络ocr › halcon基于mlp神经网络分类器的OCR字符识别 |
halcon基于mlp神经网络分类器的OCR字符识别
|
OCR字符识别常用流程如下： 1.读取图像 2.预处理 3.图像分割 4.创建字符标识关联图像区域形成.trf文件 5.创建mlp神经网络分类器create_ocr_class_mlp，然后训练 6.保存.omc文件 7.识别 按照如上的流程，通过一张图实现二十六个字母的训练，在另一张图上实现字母的识别，代码部分包含详细的注释，直接贴上代码如下： dev_close_window () *读图 read_image (Image, 'C:/Users/Administrator/Desktop/字母/81i58PICZ89.jpg') get_image_size (Image, Width, Height) dev_open_window (0, 0, Width, Height, 'black', WindowHandle) dev_display (Image) *字符分割 rgb1_to_gray (Image, GrayImage) threshold (GrayImage, Regions, 50, 200) connection (Regions, ConnectedRegions) sort_region (ConnectedRegions, SortedRegions, 'character', 'true', 'row') count_obj (SortedRegions, Number) *逐个显示确定顺序 for Index := 1 to Number by 1 dev_clear_window () select_obj (SortedRegions, ObjectSelected, Index) dev_display (ObjectSelected) stop () endfor *字符标识 word:= ['A','B','C','D','E','F','G','H','I','J','K','L','M','N','O','P','Q','R','S','T','U','V','W','X','Y','Z'] *创建训练文件 TrainFile:='D:\\words_A_Z.trf' *将图像区域与字符标识关联，保存到训练文件 write_ocr_trainf (SortedRegions, Image, word, TrainFile) *创建OMC文件 FontFlie:='D:\\FontA_Z.omc' *读取训练文件 read_ocr_trainf_names (TrainFile, CharacterNames, CharacterCount) *创建神经网络分类器mlp create_ocr_class_mlp (8, 10, 'constant', 'default', CharacterNames, 80, 'none', 10, 42, OCRHandle) *训练 trainf_ocr_class_mlp (OCRHandle, TrainFile, 200, 1, 0.01, Error, ErrorLog) *保存训练结果 write_ocr_class_mlp (OCRHandle, FontFlie) clear_ocr_class_mlp (OCRHandle) *文字识别 dev_close_window () read_image (Image1, 'C:/Users/Administrator/Desktop/字母/3.jpg') get_image_size (Image1, Width, Height) dev_open_window (0, 0, Width, Height, 'black', WindowHandle) dev_display (Image1) text_line_orientation (ObjectSelected, Image1, 25, -0.523599, 0.523599, OrientationAngle) rotate_image (Image1, ImageRotate, OrientationAngle, 'constant') *字符分割 rgb1_to_gray (ImageRotate, GrayImage1) threshold (GrayImage1, Regions2, 0, 55) connection (Regions2, ConnectedRegions1) select_shape (ConnectedRegions1, SelectedRegions, ['area','height'], 'and', [63.54,18.089], [249.07,20]) sort_region (SelectedRegions, SortedRegions1, 'character', 'true', 'row') count_obj (SortedRegions1, Number1) *识别 read_ocr_class_mlp (FontFlie, OCRHandle1) do_ocr_multi_class_mlp (SortedRegions1, GrayImage1, OCRHandle1, Class, Confidence) dev_display (Image1) for j := 1 to Number1 by 1 select_obj (SortedRegions1, ObjectSelected1, j) area_center (ObjectSelected1, Area, Row, Column) disp_message (WindowHandle, Class[j-1], 'window', Row+10, Column, 'black', 'true') endfor字符分割 
识别结果 源码和图片下载地址：点击打开链接 |
【本文地址】
今日新闻 |
推荐新闻 |