halcon字符识别与训练

版权声明：本文为博主原创文章，未经博主允许不得转载。 https://blog.csdn.net/wei_supreme/article/details/82227814

图片下载：https://weisupreme.github.io/tutorial/2018/08/08/halcon-ocr.html
  训练代码：

    *读取图片

    read_image (Image, 'D:/github/CodeRecognition/test.png')

    *反转图片，使前景为白色，便于处理

    invert_image (Image, ImageInvert)

    *使用阈值分割，提取前景

    threshold (ImageInvert, Region, 128, 255)

    *计算连通域

    connection (Region, ConnectedRegions)

    *选择符合要求的区域，即字符

    select_shape (ConnectedRegions, SelectedRegions, 'area', 'and', 150, 99999)

    *排序

    sort_region (SelectedRegions, SortedRegions, 'first_point', 'true', 'column')

    *训练****************

    *计算字符个数

    count_obj (SortedRegions, Number)

    *创建训练文件******

    *标签

    words:=['2','0','1','7']

    TrainFile:='D:/github/CodeRecognition/testWords.trf'

    dev_set_check('~give_error') 

    delete_file(TrainFile) 

    dev_set_check('~give_error')

    *生成训练文件

    for i:=1 to Number by 1 

        select_obj(SortedRegions, SingleWord, i) 

        append_ocr_trainf(SingleWord,Image,words[i-1],TrainFile) 

    endfor

    *训练ocr

    FontFile:='D:/github/CodeRecognition/testWords.omc'

    *读取训练文件

    read_ocr_trainf_names(TrainFile, CharacterNames, CharacterCount)

    NumHidden:=400

    *自己创建神经网络分类器

    *create_ocr_class_mlp (10, 20, 'constant', 'default', CharacterNames, NumHidden, 'none', 10, 42, OCRHandle)

    *这里采用halcon预训练模型；第二次及以后训练先将文件名改为FontFile，调用已训练的参数继续训练

    read_ocr_class_mlp('HandWritten_0-9.omc', OCRHandle)

    *训练

    trainf_ocr_class_mlp (OCRHandle, TrainFile, 200, 1, 0.01, Error, ErrorLog)

    *保存参数到自己命名的文件

    write_ocr_class_mlp(OCRHandle, FontFile) 

    *释放内存

    clear_ocr_class_mlp(OCRHandle)

  识别代码：

    read_image (Image, 'D:/github/CodeRecognition/test.png')

    invert_image (Image, ImageInvert)

    threshold (ImageInvert, Region, 128, 255)

    connection (Region, ConnectedRegions)

    select_shape (ConnectedRegions, SelectedRegions, 'area', 'and', 150, 99999)

    sort_region (SelectedRegions, SortedRegions, 'first_point', 'true', 'column')

    *计算每一个字符区域中心

    area_center (SortedRegions, Area, Row, Column)

    FontFile:='D:/github/CodeRecognition/testWords.omc'

    *读取已训练好的参数

    read_ocr_class_mlp(FontFile, OCRHandle)

    *识别

    do_ocr_multi_class_mlp (SortedRegions, Image, OCRHandle, RecNum, Confidence)

    *显示在屏幕上

    set_display_font (3600, 27, 'mono', 'true', 'false')

    for i := 0 to |RecNum| - 1 by 1

        disp_message (3600, RecNum[i], 'image', Row[i],Column[i], 'white', 'false')

    endfor