Halcon例子难点解释

• This example shows how to use the component-based matching in order

• to locate all instances of a dip switch module that consists of several

• components. The body of the module and each of the twelve switches

• represent a separate component. The relative movements (relations) are

• automatically trained by using twelve training images each of which shows

• another switch turned on while the remaining switches are turned off. Thus,

• the training images show all possible relative movements of the components,

• and hence cover all possible relative movements in a run-time image.

• After creating the component model all instances of the dip switch module are

• found in a run-time image. Finally, the found instances are horizontally aligned

• in order to decide whether the switches are turned on or off.
  dev_update_off ()
  dev_set_draw (‘margin’)
  read_image (ModelImage, ‘dip_switch/dip_switch_model’)
  get_image_size (ModelImage, Width, Height)
  dev_close_window ()
  dev_open_window (0, 0, Width, Height, ‘black’, WindowHandle)
  set_display_font (WindowHandle, 14, ‘mono’, ‘true’, ‘false’)

• Define the initial components
  dev_display (ModelImage)
  *截取初始组件（不会变化）
  gen_rectangle2 (InitialComponents, 265, 138, -0.02, 23, 13)
  gen_rectangle2 (InitialComponent, 342, 286, -0.02, 168, 13)
  concat_obj (InitialComponents, InitialComponent, InitialComponents)
  for I := 0 to 11 by 1
  gen_rectangle2 (InitialComponent, 311 + I * 0.5, 128 + I * 28.2, 0, 10, 12)
  concat_obj (InitialComponents, InitialComponent, InitialComponents)
  endfor
  dev_set_colored (12)
  dev_display (ModelImage)
  dev_display (InitialComponents)
  dev_set_color (‘yellow’)
  disp_message (WindowHandle, ‘Model image and’, ‘image’, 20, 20, ‘yellow’, ‘false’)
  disp_message (WindowHandle, ‘input regions describing the initial components’, ‘image’, 40, 20, ‘yellow’, ‘false’)
  disp_message (WindowHandle, ‘Press mouse button to continue’, ‘window’, 450, 20, ‘yellow’, ‘false’)
  get_mbutton (WindowHandle, Row1, Column1, Button1)

• Get the training images，获取所有情况的训练图像并组合到一起
  gen_empty_obj (TrainingImages)
  for I := 1 to 12 by 1
  read_image (TrainingImage, ‘dip_switch/dip_switch_training_’ + I$‘02’)
  concat_obj (TrainingImages, TrainingImage, TrainingImages)
  dev_display (TrainingImage)
  disp_message (WindowHandle, ‘Training image ’ + I + ’ of 12’, ‘image’, 20, 20, ‘yellow’, ‘false’)
  disp_message (WindowHandle, ‘Press mouse button to continue’, ‘window’, 450, 20, ‘yellow’, ‘false’)
  get_mbutton (WindowHandle, Row2, Column2, Button2)
  endfor

• The component-based matching uses the contours of the components

• as features for the matching. In order to make sure that only significant

• features are involved in the matching the parameters ContrastLow,

• ContrastHigh, and MinSize can be passed to the training. To inspect the

• effect of different values for these three parameters, and hence to find

• the optimum values for a certain application, the following two lines of

• code can be used:

• add_channels (InitialComponents, ModelImage, ModelImageInitComp)

• gen_initial_components (ModelImageInitComp, InitialComponentEdges, 45, 45, 30, ‘connection’, [], [])

• Extract the model components and train the relations

• (This may take a few minutes!)
  dev_display (ModelImage)
  disp_message (WindowHandle, ‘Train the model components…’, ‘image’, 20, 20, ‘yellow’, ‘false’)
  disp_message (WindowHandle, ‘(This may take a few minutes)’, ‘image’, 40, 20, ‘yellow’, ‘false’)
  dev_set_check (’~give_error’)
  dev_error_var (Error, 1)
  read_training_components (‘train_dip_switch.ct’, ComponentTrainingID)
  dev_error_var (Error, 0)
  dev_set_check (‘give_error’)
  *判断是否已经存在训练文件，如果没有则重新训练一次

if (Error != H_MSG_TRUE)
train_model_components (ModelImage, InitialComponents, TrainingImages, ModelComponents, 45, 45, 30, 0.95, -1, -1, rad(20), ‘speed’, ‘rigidity’, 0.2, 0.5, ComponentTrainingID)
* Ignore writing errors as they are not critical for this example
dev_set_check (’~give_error’)
write_training_components (ComponentTrainingID, ‘train_dip_switch.ct’)
dev_set_check (‘give_error’)
endif
dev_set_color (‘yellow’)
dev_display (ModelImage)
disp_message (WindowHandle, ‘Result of the training:’, ‘image’, 20, 20, ‘yellow’, ‘false’)
disp_message (WindowHandle, ’ 1) Final model components’, ‘image’, 40, 20, ‘yellow’, ‘false’)
dev_set_colored (12)
get_training_components (ModelComponents, ComponentTrainingID, ‘model_components’, ‘model_image’, ‘false’, RowRef, ColumnRef, AngleRef, ScoreRef)
dev_display (ModelComponents)
disp_message (WindowHandle, ‘Press mouse button to continue’, ‘window’, 450, 20, ‘yellow’, ‘false’)
get_mbutton (WindowHandle, Row3, Column3, Button3)
count_obj (ModelComponents, NumComp)
for I := 0 to NumComp - 1 by 1
dev_display (ModelImage)
disp_message (WindowHandle, ‘Result of the training:’, ‘image’, 20, 20, ‘yellow’, ‘false’)
disp_message (WindowHandle, ’ 1) Final model components’, ‘image’, 40, 20, ‘yellow’, ‘false’)
disp_message (WindowHandle, ’ 2) Relations with respect to component ’ + I, ‘image’, 60, 20, ‘yellow’, ‘false’)
dev_set_colored (12)
*获取组件的相对关系
get_component_relations (Relations, ComponentTrainingID, I, ‘model_image’, Row, Column, Phi, Length1, Length2, AngleStart, AngleExtent)
dev_display (Relations)
disp_message (WindowHandle, ‘Press mouse button to continue’, ‘window’, 450, 20, ‘yellow’, ‘false’)
get_mbutton (WindowHandle, Row4, Column4, Button4)
endfor

• Add small tolerances to the automatically derived relations in order
• to receive a higher robustness.
  *稍微休整相对关系，加大一点公差
  modify_component_relations (ComponentTrainingID, ‘all’, ‘all’, 0, rad(4))
• Create the component model based on the training result.
• The minimum score for the model components should be set individually.
• On the one hand the print on the module may be occluded, which requires
• a low minimum score. On the other hand the switches must be distinguished
• from from the background, which requires to set the corresponding minimum
• score values to higher values.
  MinScoreComp := gen_tuple_const(13,0.8)
  MinScoreComp[0] := 0.4
  dev_display (ModelImage)
  disp_message (WindowHandle, ‘Create the component model…’, ‘image’, 20, 20, ‘yellow’, ‘false’)
• Automatically determine the optimum number of pyramid levels for the
• switches. However, for the print on the module 4 pyramid levels
• instead of the automatically determined 3 levels are used to speed up
• the search.
  NumLevelsComp := gen_tuple_const(13,0)
  NumLevelsComp[0] := 4
  *注意，最后用到的是 ComponentModelID，而不是ComponentTrainingID。
  create_trained_component_model (ComponentTrainingID, 0, rad(360), 10, MinScoreComp, NumLevelsComp, ‘auto’, ‘none’, ‘use_polarity’, ‘false’, ComponentModelID, RootRanking)
  get_component_model_tree (Tree, Relations, ComponentModelID, RootRanking, ‘model_image’, StartNode, EndNode, Row, Column, Phi, Length1, Length2, AngleStart, AngleExtent)
  dev_display (ModelImage)
  disp_message (WindowHandle, ‘Component model’, ‘image’, 20, 20, ‘yellow’, ‘false’)
  dev_set_colored (12)
  dev_display (Tree)
  dev_display (Relations)
  disp_message (WindowHandle, ‘Press mouse button to continue’, ‘window’, 450, 20, ‘yellow’, ‘false’)
  get_mbutton (WindowHandle, Row5, Column5, Button5)
  wait_seconds (0.5)
  clear_training_components (ComponentTrainingID)
• Find the component model in a run-time image
  ImgNo := 1
  Button := 0
  while (Button != 1)
  read_image (SearchImage, ‘dip_switch/dip_switch_’ + ImgNo ′ 02 d ′ ) c o u n t s e c o n d s ( S e c o n d s 1 ) ∗ 参 数 解 释 ， M o d e l S t a r t ： 每 一 个 组 件 模 板 开 始 索 引 也 就 是 说 ， R o w C o m p [ M o d e l S t a r t [ 0 ] ]   R o w C o m p [ M o d e l S t a r t [ 1 ] ] , C o l u m n C o m p [ M o d e l E n d [ 0 ] ]   C o l u m n C o m p [ M o d e l E n d [ 1 ] ] 是 第 一 组 组 件 的 中 心 坐 标 ， 这 其 中 包 括 M o